CN114007132A - Anomaly detection method, device and computer-readable storage medium - Google Patents
Anomaly detection method, device and computer-readable storage medium Download PDFInfo
- Publication number
- CN114007132A CN114007132A CN202010739174.5A CN202010739174A CN114007132A CN 114007132 A CN114007132 A CN 114007132A CN 202010739174 A CN202010739174 A CN 202010739174A CN 114007132 A CN114007132 A CN 114007132A
- Authority
- CN
- China
- Prior art keywords
- moment
- video playing
- playing effect
- feature matrix
- positive integer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 65
- 230000000694 effects Effects 0.000 claims abstract description 153
- 239000011159 matrix material Substances 0.000 claims abstract description 153
- 230000002159 abnormal effect Effects 0.000 claims abstract description 51
- 238000013528 artificial neural network Methods 0.000 claims abstract description 34
- 230000005856 abnormality Effects 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000000034 method Methods 0.000 claims abstract description 11
- 230000000875 corresponding effect Effects 0.000 claims description 42
- 230000015654 memory Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 12
- 230000006870 function Effects 0.000 claims description 10
- 230000002596 correlated effect Effects 0.000 claims description 6
- 239000000126 substance Substances 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 4
- 238000010586 diagram Methods 0.000 description 16
- 238000004590 computer program Methods 0.000 description 5
- 238000012423 maintenance Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 239000013598 vector Substances 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/44—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs
- H04N21/44008—Processing of video elementary streams, e.g. splicing a video clip retrieved from local storage with an incoming video stream or rendering scenes according to encoded video stream scene graphs involving operations for analysing video streams, e.g. detecting features or characteristics in the video stream
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Computer Networks & Wireless Communication (AREA)
- Databases & Information Systems (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The disclosure provides an anomaly detection method and device and a computer readable storage medium, and relates to the technical field of information. The abnormality detection method includes: respectively collecting various video playing effect indexes at various moments; respectively generating a feature matrix of the video playing effect indexes at each moment according to each video playing effect index at each moment; processing the characteristic matrix by using a pre-trained neural network to obtain a reconstructed characteristic matrix; and detecting whether the video playing effect is abnormal or not according to the characteristic matrix and the reconstructed characteristic matrix. The method and the device can accurately and automatically detect whether the video playing effect is abnormal or not.
Description
Technical Field
The present disclosure relates to the field of information technologies, and in particular, to an anomaly detection method and apparatus, and a computer-readable storage medium.
Background
The user base number of the interactive network television (IPTV for short) service is huge, and the watching experience of the user is sensitive to the change of the running condition of the network equipment. Therefore, at present, a technical scheme capable of accurately and automatically detecting the abnormal video playing effect of the IPTV is urgently needed, so as to find that the watching experience of the IPTV user is reduced and generate a related early warning of the abnormal video playing effect in time.
Disclosure of Invention
The technical problem solved by the present disclosure is how to accurately and automatically detect whether the video playing effect is abnormal.
According to an aspect of the embodiments of the present disclosure, there is provided an abnormality detection method including: respectively collecting various video playing effect indexes at various moments; respectively generating a feature matrix of the video playing effect indexes at each moment according to each video playing effect index at each moment; processing the characteristic matrix by using a pre-trained neural network to obtain a reconstructed characteristic matrix; and detecting whether the video playing effect is abnormal or not according to the characteristic matrix and the reconstructed characteristic matrix.
In some embodiments, generating the feature matrix of the video playing effect indicators at each time according to the video playing effect indicators at each time includes: and generating a feature matrix of the first moment according to the video playing effect indexes from the first moment to the second moment, wherein the difference between the first moment and the second moment is a preset positive integer, and the first moment is any one moment in the moments.
In some embodiments, generating the feature matrix at the first time according to the video playing effect indicators from the first time to the second time includes: generating a feature matrix corresponding to a target positive integer at a first moment according to each video playing effect index from the first moment to a target second moment, wherein the target positive integer is any element in a value set of a preset positive integer, the target second moment is any element in a value set of the second moment, and the difference between the first moment and the target second moment is the target positive integer; and taking the characteristic matrixes respectively corresponding to the target positive integers at the first moment as the characteristic matrixes at the first moment.
In some embodiments, the value of the jth row and jth column element in the feature matrix corresponding to the target positive integer at the first time is positively correlated with the weighted sum of the product of the ith video playing effect index and the jth video playing effect index from the first time to the target second time.
In some embodiments, the feature matrix corresponding to the target positive integer e at time t is generated using the following formula:
wherein the content of the first and second substances,
is the element of the ith row and the jth column in the characteristic matrix corresponding to the target positive integer e at the time t, s is the accumulated mark, wsIs a weighting coefficient that is inversely related to s,
is the ith video playing effect index at the time t-s,
and k is the number of elements in the value set of the preset positive integer, and is the j-th video playing effect index at the moment of t-s.
In some embodiments, the anomaly detection method further comprises: and training the neural network by using the sample characteristic matrix of the video playing effect index at each moment, so that the neural network outputs a reconstruction characteristic matrix corresponding to the sample characteristic matrix, wherein a loss function adopted in the training of the neural network is the sum of L2 norm errors between the sample characteristic matrix at each moment and the corresponding reconstruction characteristic matrix.
In some embodiments, detecting whether the video playing effect is abnormal according to the feature matrix and the reconstructed feature matrix includes: taking an L2 norm error between the feature matrix at the first moment and the reconstructed feature matrix as a first error, wherein the first moment is any one of the moments; detecting that the video playing effect at the first moment is abnormal under the condition that the first error is greater than a first threshold value; and under the condition that the first error is not larger than the first threshold, detecting that the video playing effect at the first moment is not abnormal.
In some embodiments, the anomaly detection method further comprises: after the video playing effect at the first moment is detected to be abnormal, taking an L2 norm error between the ith row of the feature matrix at the first moment and the ith row of the reconstructed feature matrix as a second error; under the condition that the second error is larger than a second threshold value, detecting that the video playing effect is abnormal due to the ith video playing effect index at the first moment; and under the condition that the second error is not greater than the second threshold value, detecting that the video playing effect is not abnormal due to the ith video playing effect index at the first moment.
According to another aspect of the embodiments of the present disclosure, there is provided an abnormality detection apparatus including: the index acquisition module is configured to respectively acquire various video playing effect indexes at various moments; the matrix generation module is configured to respectively generate a characteristic matrix of the video playing effect indexes at each moment according to each video playing effect index at each moment; the matrix processing module is configured to process the characteristic matrix by utilizing a pre-trained neural network to obtain a reconstructed characteristic matrix; and the abnormity detection module is configured to detect whether the video playing effect is abnormal or not according to the characteristic matrix and the reconstructed characteristic matrix.
In some embodiments, the matrix generation module is configured to: and generating a feature matrix of the first moment according to the video playing effect indexes from the first moment to the second moment, wherein the difference between the first moment and the second moment is a preset positive integer, and the first moment is any one moment in the moments.
In some embodiments, the matrix generation module is configured to: generating a feature matrix corresponding to a target positive integer at a first moment according to each video playing effect index from the first moment to a target second moment, wherein the target positive integer is any element in a value set of a preset positive integer, the target second moment is any element in a value set of the second moment, and the difference between the first moment and the target second moment is the target positive integer; and taking the characteristic matrixes respectively corresponding to the target positive integers at the first moment as the characteristic matrixes at the first moment.
In some embodiments, the value of the jth row and jth column element in the feature matrix corresponding to the target positive integer at the first time is positively correlated with the weighted sum of the product of the ith video playing effect index and the jth video playing effect index from the first time to the target second time.
In some embodiments, the matrix generation module is configured to: generating a characteristic matrix corresponding to the target positive integer e at t time by adopting the following formula:
wherein the content of the first and second substances,
is the element of the ith row and the jth column in the characteristic matrix corresponding to the target positive integer e at the time t, s is the accumulated mark, wsIs a weighting coefficient that is inversely related to s,
is the ith video playing effect index at the time t-s,
j-th video broadcast at time t-sAnd (5) putting the effect index, wherein k is the number of elements in the value set of the preset positive integer.
In some embodiments, the abnormality detection apparatus further includes: and the network training module is configured to train the neural network by using the sample characteristic matrix of the video playing effect index at each moment, so that the neural network outputs a reconstruction characteristic matrix corresponding to the sample characteristic matrix, and the loss function adopted in the training of the neural network is the sum of L2 norm errors between the sample characteristic matrix at each moment and the corresponding reconstruction characteristic matrix.
In some embodiments, the anomaly detection module is configured to: taking an L2 norm error between the feature matrix at the first moment and the reconstructed feature matrix as a first error, wherein the first moment is any one of the moments; detecting that the video playing effect at the first moment is abnormal under the condition that the first error is greater than a first threshold value; and under the condition that the first error is not larger than the first threshold, detecting that the video playing effect at the first moment is not abnormal.
In some embodiments, the anomaly detection apparatus further comprises an index detection module configured to: after the video playing effect at the first moment is detected to be abnormal, taking an L2 norm error between the ith row of the feature matrix at the first moment and the ith row of the reconstructed feature matrix as a second error; under the condition that the second error is larger than a second threshold value, detecting that the video playing effect is abnormal due to the ith video playing effect index at the first moment; and under the condition that the second error is not greater than the second threshold value, detecting that the video playing effect is not abnormal due to the ith video playing effect index at the first moment.
According to still another aspect of the embodiments of the present disclosure, there is provided an abnormality detection apparatus including: a memory; and a processor coupled to the memory, the processor configured to perform the foregoing anomaly detection method based on instructions stored in the memory.
According to still another aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the instructions when executed by a processor implement the foregoing anomaly detection method.
The method and the device can accurately and automatically detect whether the video playing effect is abnormal or not.
Other features of the present disclosure and advantages thereof will become apparent from the following detailed description of exemplary embodiments thereof, which proceeds with reference to the accompanying drawings.
Drawings
In order to more clearly illustrate the embodiments of the present disclosure or technical solutions in the related art, the drawings required to be used in the description of the embodiments or the related art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and for those skilled in the art, other drawings may be obtained according to the drawings without inventive exercise.
Fig. 1 illustrates a flow diagram of an anomaly detection method of some embodiments of the present disclosure.
Fig. 2 shows a schematic diagram of a pre-trained neural network.
FIG. 3 is a flow chart illustrating an anomaly detection method according to further embodiments of the present disclosure.
Fig. 4 shows a schematic structural diagram of an abnormality detection apparatus according to some embodiments of the present disclosure.
Fig. 5 is a schematic structural diagram of an abnormality detection apparatus according to another embodiment of the present disclosure.
Detailed Description
The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure, and it is obvious that the described embodiments are only a part of the embodiments of the present disclosure, and not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the disclosure, its application, or uses. All other embodiments, which can be derived by a person skilled in the art from the embodiments disclosed herein without making any creative effort, shall fall within the protection scope of the present disclosure.
The traditional IPTV user experience abnormity detection method generally adopts a threshold alarm mode, namely, thresholds are respectively set for various video playing effect indexes, and IPTV user experience abnormity is detected under the condition that the video playing effect indexes exceed the thresholds. This approach has the following problems: on one hand, the traditional method is very sensitive to the setting of the threshold value, and the noise point and the burr point caused by the frequent fluctuation of each video playing effect index are easy to generate false detection and false alarm; on the other hand, the conventional method only sets a plurality of thresholds respectively and gives an alarm for different video playing effect indexes aiming at variable service scenes, and lacks comprehensive consideration for the video playing effect indexes.
In view of this, the present disclosure provides an anomaly detection method for IPTV user experience, which can accurately and automatically detect whether an anomaly occurs in a video playing effect.
Some embodiments of the disclosed anomaly detection method are first described in conjunction with FIG. 1.
Fig. 1 illustrates a flow diagram of an anomaly detection method of some embodiments of the present disclosure. As shown in fig. 1, the present embodiment includes steps S101 to S104.
In step S101, each video playing effect index at each time is collected.
Firstly, multi-dimensional video playing effect indexes at each time are collected, such as playing card pause user duty ratio after normalization processing, playing card pause packet loss user number after normalization processing, control response slow user number after normalization processing, set top box loading time after normalization processing and the like. Suppose that n video playing effect indexes at each time are collected at T (for example, taking a value of 24) times in total.
In step S102, a feature matrix of the video playing effect index at each time is generated according to each video playing effect index at each time.
When the feature matrix is generated, the playing effect indexes of each video at each moment can be comprehensively considered from the index dimension and the time dimension. Assume that a first time T (T < T) is any one of the times T, and a second time T-E is a time before the first time, i.e., the difference between the first time and the second time is a preset positive integer E. Then, the feature matrix at the first time may be generated according to the video playing effect indexes from the first time t to the second time t-E. It should be understood by those skilled in the art that when the first time is the first several times of the T times, the second time T-E may not belong to each of the T times, and at this time, data padding may be performed on several times before the T times to obtain various video playing effect indexes at various times from the second time T-E to the first time T.
Specifically, a feature matrix corresponding to the target positive integer e at the first time t may be generated according to various video playing effect indexes from the first time t to the target second time t-e. The target positive integer E is any element in a value set of a preset positive integer E, the target second time t-E is any element in the value set of the second time t-E, and the difference between the first time t and the target second time t-E is the target positive integer E. Then, the feature matrices corresponding to the target positive integers in E at the first time t may be used as the feature matrices at the first time t.
In some embodiments, the value of the jth row and jth column element in the feature matrix corresponding to the target positive integer at the first time is positively correlated with the weighted sum of the product of the ith video playing effect index and the jth video playing effect index from the first time to the target second time. For example, the feature matrix corresponding to the target positive integer e at time t may be generated by using the following formula:
wherein the content of the first and second substances,
is the element of the ith row and the jth column in the characteristic matrix corresponding to the target positive integer e at the time t, s is the accumulated mark, wsFor the weighting coefficient having negative correlation with s (i.e. the closer the target second time is, the weaker the correlation of the video playing effect index between the target second time and the first time is, w can be taken asA value of 0.3),
is the ith video playing effect index at the time t-s,
and k is the number of elements in the value set of the preset positive integer, and is the j-th video playing effect index at the moment of t-s. Taking different values from E for E (for example, E can be respectively taken as 1, 2, 4 and 6), and finally obtaining T feature matrices with the size of n × n channels as k, wherein the T feature matrices with the size of n × n channels as k can reflect time related information and index related information of each video playing effect index at T moments.
In step S103, the feature matrix is processed by using a pre-trained neural network to obtain a reconstructed feature matrix.
Fig. 2 shows a schematic diagram of a pre-trained neural network. As shown in FIG. 2, the neural Network may be, for example, a self-coding neural Network, the structure of the self-coding neural Network itself does not belong to the invention of the present disclosure, and specifically, the papers "Multi-Scale spatial statistical Conv-LSTM Network for Video sales Detection", "Attention in volumetric LSTM for Gesture Recognition", "CNN-LSTM model and application based on Attention machine", and "Chinese sign language Recognition based on space-time Attention Network" may be referred to, and are not detailed herein. In this embodiment, the input data from the encoded neural network is a feature matrix with a time length of T, k channels and a size of n × n. The encoder part of the self-coding neural network comprises a module a and a module b, and the decoder part can comprise a module c. Performing convolution calculation on the feature matrix by each convolution layer in the module a, and pouring the output of each convolution layer in the module a into the module b; each convolution-long short-term memory network (ConvLSTM) in block b can use a mechanism of attention to output feature vectors from the hidden layer and inject them into block c. The c module can perform deconvolution (Deconv for short) and merging (Concat for short) operations on the feature vectors output by the b module, and finally obtains a reconstructed feature matrix.
For example, after training the neural network by using the sample feature matrix of the video playing effect index at each time (8000 sample feature matrices in total, and one batch of 256 sample feature matrices), the neural network may output a reconstructed feature matrix corresponding to the sample feature matrix. The loss function used in training the neural network is the sum of the L2 norm errors between the sample feature matrix at each time instant and the corresponding reconstructed feature matrix.
In step S104, it is detected whether the video playing effect is abnormal or not according to the feature matrix and the reconstructed feature matrix.
When detecting whether the video playing effect is abnormal, an L2 norm error between the feature matrix at the first time and the reconstructed feature matrix may be used as the first error, where the first time is any one of the respective times. Then, it is determined whether the first error is greater than a first threshold.
Detecting that the video playing effect at the first moment is abnormal under the condition that the first error is greater than a first threshold value; and under the condition that the first error is not larger than the first threshold, detecting that the video playing effect at the first moment is not abnormal.
E.g. first error at time t
The calculation formula of (c) may be:
wherein m'ijeThe number of channels in the ith row and the jth column in the reconstructed feature matrix is e. Those skilled in the art will appreciate that the first threshold value can be flexibly set according to actual requirements. If it is
If the value is larger than the first threshold, it can be detected that the video playing effect at the time t is abnormal.
The embodiment constructs the characteristic matrix with richer information quantity in the index dimension and the time dimension, can fully reflect the time associated information of the video playing effect index and the associated information between the indexes, and can effectively reduce the interference of noise points and burr points on detecting whether the video playing effect is abnormal or not. Then, whether the video playing effect is abnormal or not can be detected accurately and automatically based on the self-coding neural network, and the judgment is carried out without setting independent threshold values respectively for different video playing effect indexes, so that the method has the characteristics of convenience, high efficiency, flexibility, reliability, strong generalization and wide application range.
Further embodiments of the disclosed anomaly detection method are described below in conjunction with FIG. 3.
FIG. 3 is a flow chart illustrating an anomaly detection method according to further embodiments of the present disclosure. As shown in fig. 3, the present embodiment includes steps S305 to S308.
In step S305, after detecting that the video playing effect is abnormal at the first time, an L2 norm error between the ith row of the feature matrix at the first time and the ith row of the reconstructed feature matrix is taken as a second error.
In step S306, it is determined whether the second error is greater than a second threshold.
If the second error is greater than the second threshold, executing step S307, that is, it is detected that the video playing effect is abnormal due to the ith video playing effect index at the first time;
if the second error is not greater than the second threshold, step S308 is executed, that is, it is detected that the video playing effect index of the ith item at the first time does not cause the video playing effect to be abnormal.
E.g. second error at time t
The calculation formula of (c) may be:
if it is
If the value is larger than the second threshold, it can be detected that the video playing effect is abnormal due to the ith video playing effect index at the time t.
The video playing effect index which causes the video playing effect to be abnormal can be accurately, conveniently and efficiently detected by the embodiment. In an actual application scene, the fault location module is combined with other network equipment fault location modules, so that the fault repair work order can be automatically distributed to operation and maintenance personnel guaranteed by the IPTV network, the working cost of manual analysis is saved, the automatic operation and maintenance of the IPTV network are realized, and the normal operation and user experience of the IPTV service are guaranteed.
Some embodiments of the disclosed anomaly detection apparatus are described below in conjunction with FIG. 4.
Fig. 4 shows a schematic structural diagram of an abnormality detection apparatus according to some embodiments of the present disclosure. As shown in fig. 4, the abnormality detection device 40 in the present embodiment includes: an index collection module 401 configured to collect each video playing effect index at each time respectively; a matrix generating module 402, configured to generate a feature matrix of the video playing effect index at each time according to each video playing effect index at each time; a matrix processing module 403 configured to process the feature matrix by using a pre-trained neural network to obtain a reconstructed feature matrix; and an anomaly detection module 404 configured to detect whether an anomaly occurs in the video playing effect according to the feature matrix and the reconstructed feature matrix.
In some embodiments, the matrix generation module 402 is configured to: and generating a feature matrix of the first moment according to the video playing effect indexes from the first moment to the second moment, wherein the difference between the first moment and the second moment is a preset positive integer, and the first moment is any one moment in the moments.
In some embodiments, the matrix generation module 402 is configured to: generating a feature matrix corresponding to a target positive integer at a first moment according to each video playing effect index from the first moment to a target second moment, wherein the target positive integer is any element in a value set of a preset positive integer, the target second moment is any element in a value set of the second moment, and the difference between the first moment and the target second moment is the target positive integer; and taking the characteristic matrixes respectively corresponding to the target positive integers at the first moment as the characteristic matrixes at the first moment.
In some embodiments, the value of the jth row and jth column element in the feature matrix corresponding to the target positive integer at the first time is positively correlated with the weighted sum of the product of the ith video playing effect index and the jth video playing effect index from the first time to the target second time.
In some embodiments, the matrix generation module is configured to: generating a characteristic matrix corresponding to the target positive integer e at t time by adopting the following formula:
wherein the content of the first and second substances,
is the element of the ith row and the jth column in the characteristic matrix corresponding to the target positive integer e at the time t, s is the accumulated mark, wsIs a weighting coefficient that is inversely related to s,
is the ith video playing effect index at the time t-s,
and k is the number of elements in the value set of the preset positive integer, and is the j-th video playing effect index at the moment of t-s.
In some embodiments, the anomaly detection apparatus 40 further includes a network training module 400 configured to train the neural network by using the sample feature matrix of the video playing effect index at each time, so that the neural network outputs a reconstructed feature matrix corresponding to the sample feature matrix, and the loss function used in training the neural network is the sum of the L2 norm errors between the sample feature matrix at each time and the corresponding reconstructed feature matrix.
In some embodiments, the anomaly detection module 404 is configured to: taking an L2 norm error between the feature matrix at the first moment and the reconstructed feature matrix as a first error, wherein the first moment is any one of the moments; detecting that the video playing effect at the first moment is abnormal under the condition that the first error is greater than a first threshold value; and under the condition that the first error is not larger than the first threshold, detecting that the video playing effect at the first moment is not abnormal.
The embodiment constructs the characteristic matrix with richer information quantity in the index dimension and the time dimension, can fully reflect the time associated information of the video playing effect index and the associated information between the indexes, and can effectively reduce the interference of noise points and burr points on detecting whether the video playing effect is abnormal or not. Then, whether the video playing effect is abnormal or not can be detected accurately and automatically based on the self-coding neural network, and the judgment is carried out without setting independent threshold values respectively for different video playing effect indexes, so that the method has the characteristics of convenience, high efficiency, flexibility, reliability, strong generalization and wide application range.
In some embodiments, the anomaly detection apparatus 40 further includes an indicator detection module 405 configured to: after the video playing effect at the first moment is detected to be abnormal, taking an L2 norm error between the ith row of the feature matrix at the first moment and the ith row of the reconstructed feature matrix as a second error; under the condition that the second error is larger than a second threshold value, detecting that the video playing effect is abnormal due to the ith video playing effect index at the first moment; and under the condition that the second error is not greater than the second threshold value, detecting that the video playing effect is not abnormal due to the ith video playing effect index at the first moment.
The video playing effect index which causes the video playing effect to be abnormal can be accurately, conveniently and efficiently detected by the embodiment. In an actual application scene, the fault location module is combined with other network equipment fault location modules, so that the fault repair work order can be automatically distributed to operation and maintenance personnel guaranteed by the IPTV network, the working cost of manual analysis is saved, the automatic operation and maintenance of the IPTV network are realized, and the normal operation and user experience of the IPTV service are guaranteed.
Further embodiments of the anomaly detection apparatus of the present disclosure are described below in conjunction with FIG. 5.
Fig. 5 is a schematic structural diagram of an abnormality detection apparatus according to another embodiment of the present disclosure. As shown in fig. 5, the abnormality detection device 50 of this embodiment includes: a memory 510 and a processor 520 coupled to the memory 510, the processor 520 being configured to perform the method of anomaly detection in any of the embodiments described above based on instructions stored in the memory 510.
The abnormality detection apparatus 50 may further include an input-output interface 530, a network interface 540, a storage interface 550, and the like. These interfaces 530, 540, 550 and the connections between the memory 510 and the processor 520 may be, for example, via a bus 560. The input/output interface 530 provides a connection interface for input/output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 540 provides a connection interface for various networking devices. The storage interface 550 provides a connection interface for external storage devices such as an SD card and a usb disk.
The present disclosure also includes a computer readable storage medium having stored thereon computer instructions that, when executed by a processor, implement the anomaly detection method in any of the foregoing embodiments.
The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the disclosure. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only exemplary of the present disclosure and is not intended to limit the present disclosure, so that any modification, equivalent replacement, or improvement made within the spirit and principle of the present disclosure should be included in the scope of the present disclosure.
Claims (18)
1. An anomaly detection method comprising:
respectively collecting various video playing effect indexes at various moments;
respectively generating a feature matrix of the video playing effect indexes at each moment according to each video playing effect index at each moment;
processing the characteristic matrix by utilizing a pre-trained neural network to obtain a reconstructed characteristic matrix;
and detecting whether the video playing effect is abnormal or not according to the feature matrix and the reconstructed feature matrix.
2. The anomaly detection method according to claim 1, wherein the generating a feature matrix of the video playing effect indexes at each time according to the video playing effect indexes at each time comprises:
and generating the feature matrix of the first moment according to the video playing effect indexes from the first moment to the second moment, wherein the difference between the first moment and the second moment is a preset positive integer, and the first moment is any one moment in the moments.
3. The anomaly detection method according to claim 2, wherein the generating the feature matrix at the first time according to the video playing effect indicators from the first time to the second time comprises:
generating the feature matrix corresponding to a target positive integer at a first moment according to each video playing effect index from the first moment to a target second moment, wherein the target positive integer is any element in a value set of a preset positive integer, the target second moment is any element in a value set of the second moment, and the difference between the first moment and the target second moment is the target positive integer;
and taking the feature matrix corresponding to each target positive integer at the first moment as the feature matrix at the first moment.
4. The abnormality detection method according to claim 3, wherein a value of an ith row and a jth column element in the feature matrix corresponding to the target positive integer at the first time is positively correlated with a weighted sum of a product of an ith video play effect indicator and a jth video play effect indicator from the first time to the target second time.
5. The abnormality detection method according to claim 4, wherein the feature matrix corresponding to a target positive integer e at time t is generated using the following formula:
wherein the content of the first and second substances,
is the element of the ith row and the jth column in the characteristic matrix corresponding to the target positive integer e at the time t, s is the accumulated mark, wsIs a weighting coefficient that is inversely related to s,
is the ith video playing effect index at the time t-s,
and k is the number of elements in the value set of the preset positive integer, and is the j-th video playing effect index at the moment of t-s.
6. The abnormality detection method according to claim 1, further comprising:
and training the neural network by using the sample characteristic matrix of the video playing effect index at each moment, so that the neural network outputs a reconstruction characteristic matrix corresponding to the sample characteristic matrix, wherein a loss function adopted in training the neural network is the sum of L2 norm errors between the sample characteristic matrix and the corresponding reconstruction characteristic matrix at each moment.
7. The anomaly detection method according to claim 1, wherein said detecting whether an anomaly occurs in a video playing effect according to the feature matrix and the reconstructed feature matrix comprises:
taking an L2 norm error between the feature matrix at a first moment and the reconstructed feature matrix as a first error, wherein the first moment is any one of the moments;
detecting that the video playing effect at the first moment is abnormal under the condition that the first error is larger than a first threshold value;
and under the condition that the first error is not larger than a first threshold value, detecting that the video playing effect at the first moment is not abnormal.
8. The abnormality detection method according to claim 7, further comprising:
after the video playing effect at the first moment is detected to be abnormal, taking an L2 norm error between the ith row of the feature matrix at the first moment and the ith row of the reconstructed feature matrix as a second error;
under the condition that the second error is larger than a second threshold value, detecting that the video playing effect is abnormal due to the ith video playing effect index at the first moment;
and under the condition that the second error is not greater than a second threshold value, detecting that the video playing effect is not abnormal due to the ith video playing effect index at the first moment.
9. An abnormality detection device comprising:
the index acquisition module is configured to respectively acquire various video playing effect indexes at various moments;
the matrix generation module is configured to respectively generate a characteristic matrix of the video playing effect indexes at each moment according to each video playing effect index at each moment;
the matrix processing module is configured to process the characteristic matrix by utilizing a pre-trained neural network to obtain a reconstructed characteristic matrix;
and the abnormity detection module is configured to detect whether the video playing effect is abnormal or not according to the characteristic matrix and the reconstructed characteristic matrix.
10. The anomaly detection apparatus of claim 9, wherein the matrix generation module is configured to:
and generating the feature matrix of the first moment according to the video playing effect indexes from the first moment to the second moment, wherein the difference between the first moment and the second moment is a preset positive integer, and the first moment is any one moment in the moments.
11. The anomaly detection apparatus of claim 10, wherein the matrix generation module is configured to:
generating the feature matrix corresponding to a target positive integer at a first moment according to each video playing effect index from the first moment to a target second moment, wherein the target positive integer is any element in a value set of a preset positive integer, the target second moment is any element in a value set of the second moment, and the difference between the first moment and the target second moment is the target positive integer;
and taking the feature matrix corresponding to each target positive integer at the first moment as the feature matrix at the first moment.
12. The abnormality detection device according to claim 11, wherein a value of an ith row and a jth column element in the feature matrix corresponding to the target positive integer at the first time is positively correlated with a weighted sum of a product of an ith video play effect indicator and a jth video play effect indicator from the first time to the target second time.
13. The anomaly detection apparatus of claim 12, wherein the matrix generation module is configured to: generating the feature matrix corresponding to the target positive integer e at t time by adopting the following formula:
wherein the content of the first and second substances,
is the element of the ith row and the jth column in the characteristic matrix corresponding to the target positive integer e at the time t, s is the accumulated mark, wsIs a weighting coefficient that is inversely related to s,
is the ith video playing effect index at the time t-s,
and k is the number of elements in the value set of the preset positive integer, and is the j-th video playing effect index at the moment of t-s.
14. The abnormality detection device according to claim 9, further comprising:
the network training module is configured to train the neural network by using the sample feature matrix of the video playing effect index at each moment, so that the neural network outputs a reconstruction feature matrix corresponding to the sample feature matrix, and a loss function adopted in the training of the neural network is the sum of L2 norm errors between the sample feature matrix and the corresponding reconstruction feature matrix at each moment.
15. The anomaly detection device of claim 9, wherein the anomaly detection module is configured to:
taking an L2 norm error between the feature matrix at a first moment and the reconstructed feature matrix as a first error, wherein the first moment is any one of the moments;
detecting that the video playing effect at the first moment is abnormal under the condition that the first error is larger than a first threshold value;
and under the condition that the first error is not larger than a first threshold value, detecting that the video playing effect at the first moment is not abnormal.
16. The abnormality detection device according to claim 12, further comprising an index detection module configured to:
after the video playing effect at the first moment is detected to be abnormal, taking an L2 norm error between the ith row of the feature matrix at the first moment and the ith row of the reconstructed feature matrix as a second error;
under the condition that the second error is larger than a second threshold value, detecting that the video playing effect is abnormal due to the ith video playing effect index at the first moment;
and under the condition that the second error is not greater than a second threshold value, detecting that the video playing effect is not abnormal due to the ith video playing effect index at the first moment.
17. An abnormality detection device comprising:
a memory; and
a processor coupled to the memory, the processor configured to perform the anomaly detection method of any of claims 1-8 based on instructions stored in the memory.
18. A computer readable storage medium, wherein the computer readable storage medium stores computer instructions which, when executed by a processor, implement the anomaly detection method of any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010739174.5A CN114007132A (en) | 2020-07-28 | 2020-07-28 | Anomaly detection method, device and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010739174.5A CN114007132A (en) | 2020-07-28 | 2020-07-28 | Anomaly detection method, device and computer-readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114007132A true CN114007132A (en) | 2022-02-01 |
Family
ID=79920529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010739174.5A Pending CN114007132A (en) | 2020-07-28 | 2020-07-28 | Anomaly detection method, device and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114007132A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115542172A (en) * | 2022-12-01 | 2022-12-30 | 湖北工业大学 | Power battery fault detection method, system, device and storage medium |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013150083A (en) * | 2012-01-18 | 2013-08-01 | Nippon Telegr & Teleph Corp <Ntt> | Network abnormality detection device and network abnormality detection method |
| CN104268594A (en) * | 2014-09-24 | 2015-01-07 | 中安消技术有限公司 | Method and device for detecting video abnormal events |
| CN105810213A (en) * | 2014-12-30 | 2016-07-27 | 浙江大华技术股份有限公司 | Typical abnormal sound detection method and device |
| US20180176243A1 (en) * | 2016-12-16 | 2018-06-21 | Patternex, Inc. | Method and system for learning representations for log data in cybersecurity |
| CN108509827A (en) * | 2017-02-27 | 2018-09-07 | 阿里巴巴集团控股有限公司 | The recognition methods of anomalous content and video flow processing system and method in video flowing |
| CN108965934A (en) * | 2017-05-27 | 2018-12-07 | 华为技术有限公司 | A kind of fault detection method, monitoring device and the network equipment |
| US20200064822A1 (en) * | 2018-08-27 | 2020-02-27 | Nec Laboratories America, Inc. | Unsupervised anomaly detection, diagnosis, and correction in multivariate time series data |
| CN111291411A (en) * | 2020-02-13 | 2020-06-16 | 福州大学 | Safe video anomaly detection system and method based on convolutional neural network |
-
2020
- 2020-07-28 CN CN202010739174.5A patent/CN114007132A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2013150083A (en) * | 2012-01-18 | 2013-08-01 | Nippon Telegr & Teleph Corp <Ntt> | Network abnormality detection device and network abnormality detection method |
| CN104268594A (en) * | 2014-09-24 | 2015-01-07 | 中安消技术有限公司 | Method and device for detecting video abnormal events |
| CN105810213A (en) * | 2014-12-30 | 2016-07-27 | 浙江大华技术股份有限公司 | Typical abnormal sound detection method and device |
| US20180176243A1 (en) * | 2016-12-16 | 2018-06-21 | Patternex, Inc. | Method and system for learning representations for log data in cybersecurity |
| CN108509827A (en) * | 2017-02-27 | 2018-09-07 | 阿里巴巴集团控股有限公司 | The recognition methods of anomalous content and video flow processing system and method in video flowing |
| CN108965934A (en) * | 2017-05-27 | 2018-12-07 | 华为技术有限公司 | A kind of fault detection method, monitoring device and the network equipment |
| US20200064822A1 (en) * | 2018-08-27 | 2020-02-27 | Nec Laboratories America, Inc. | Unsupervised anomaly detection, diagnosis, and correction in multivariate time series data |
| CN111291411A (en) * | 2020-02-13 | 2020-06-16 | 福州大学 | Safe video anomaly detection system and method based on convolutional neural network |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115542172A (en) * | 2022-12-01 | 2022-12-30 | 湖北工业大学 | Power battery fault detection method, system, device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111178456B (en) | Abnormal index detection method and device, computer equipment and storage medium | |
| CN110097037B (en) | Intelligent monitoring method and device, storage medium and electronic equipment | |
| US11169514B2 (en) | Unsupervised anomaly detection, diagnosis, and correction in multivariate time series data | |
| CN108563739B (en) | Weather data acquisition method and device, computer device and readable storage medium | |
| Song et al. | Multiple event detection and recognition for large-scale power systems through cluster-based sparse coding | |
| CN107329933B (en) | Fault detection method and device based on optical fiber sensing vibration signal | |
| CN111459700A (en) | Method and apparatus for diagnosing device failure, diagnostic device, and storage medium | |
| CN108427669A (en) | Abnormal behaviour monitoring method and system | |
| CN111078552A (en) | Method and device for detecting page display abnormity and storage medium | |
| KR101964412B1 (en) | Method for diagnosing anomaly log of mobile commmunication data processing system and system thereof | |
| CN113868006B (en) | Time sequence detection method and device, electronic equipment and computer storage medium | |
| US20200041988A1 (en) | Display method, display device, and program | |
| CN115698882A (en) | Abnormal modulation cause identification device, abnormal modulation cause identification method, and abnormal modulation cause identification program | |
| US20190265088A1 (en) | System analysis method, system analysis apparatus, and program | |
| CN114007132A (en) | Anomaly detection method, device and computer-readable storage medium | |
| CN116974268B (en) | Intelligent monitoring and early warning method for control system circuit | |
| CN113536658B (en) | Electromechanical equipment lightweight fault diagnosis method based on STM32 embedded processor | |
| CN113128412B (en) | Fire trend prediction method based on deep learning and fire monitoring video | |
| CN109542722A (en) | Anomaly analysis processing method, device and storage medium | |
| CN115952020A (en) | Page abnormity detection method, medium, device and computing equipment | |
| CN114117354A (en) | Method, device and equipment for detecting abnormity of time sequence data | |
| CN111581044A (en) | Cluster optimization method, device, server and medium | |
| CN112995650B (en) | Method and device for detecting video continuity of camera | |
| CN116150666B (en) | Energy storage system fault detection method and device and intelligent terminal | |
| CN115576724B (en) | Fault isolation method, device, equipment, medium and product of PIU subsystem |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |