Object recognition method, device, electronic device and storage medium

By performing multi-dimensional feature extraction and weighting of the operation information of the object to be identified in the live broadcast software, predicting whether it will continue to broadcast live, the problem of poor interaction experience between the anchor object and the audience object is solved, and the recognition accuracy and interaction effect are improved.

CN116233473BActive Publication Date: 2024-12-03BEIJING DAJIA INTERNET INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310102624.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-01-29
Publication Date
2024-12-03
Estimated Expiration
2043-01-29

AI Technical Summary

Technical Problem

The interactive experience between the anchor object and the audience object in live broadcast software is poor, resulting in the inability to form positive interaction feedback. The existing technology mainly relies on subjective experience or statistical methods to set thresholds for identification, and the identification results are not accurate enough.

Method used

By extracting the operation information of the object to be identified within N consecutive days, obtaining the first operation feature and the second operation feature, combining the live operation feature, content release feature and interaction feature, weighting processing and trend analysis are carried out to predict whether the object will continue to execute. Live broadcast operations to determine the target object.

Benefits of technology

It improves the accuracy and interactive experience of object recognition, forms positive interactive feedback, and enhances the accuracy of live broadcast operation prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116233473B_ABST
    Figure CN116233473B_ABST
Patent Text Reader

Abstract

The present disclosure relates to an object recognition method, device, electronic device and storage medium, the method comprising: extracting features of the operation information of the object to be recognized within a first time period and the operation information within a second time period, and obtaining first operation features and second operation features corresponding to the object to be recognized. The object to be recognized is an object that has not been broadcast live for N consecutive days, and the first time period is the time period before the N days; N days include the current time, and N is an integer greater than 1. Based on the first operation feature and the second operation feature, determine whether the object to be recognized will continue to perform the live broadcast operation. The method can identify potential target objects by integrating multi-dimensional features, thereby improving the interactive experience between the object to be recognized and the interactive object, and forming positive interactive feedback.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to an object recognition method, device, electronic device and storage medium. Background Art

[0002] In live broadcast software, in order to allow users of the live broadcast software to interact better, it is usually possible to identify multiple objects such as the host object and the audience object in the live broadcast software to determine the objects that match the live broadcast activity in the live broadcast software. In the related art, when performing host object identification, most of the thresholds are set based on subjective experience or statistical methods, and the host object is generally identified through the threshold, resulting in the identification result being not accurate enough and prone to generalization. As a result, in the process of interaction between the identified host object and the audience object, there is a poor interactive experience between the host object and the audience object, and the problem of being unable to form positive interactive feedback. Summary of the invention

[0003] The present disclosure provides an object recognition method, device, electronic device and storage medium to at least solve the problem in the related art that the interactive experience between the host object and the audience object is poor and positive interactive feedback cannot be formed. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided an object recognition method, the method comprising:

[0005] Extracting features from the operation information of the object to be identified within a first time period to obtain a first operation feature corresponding to the object to be identified; the object to be identified is an object that has not performed a live broadcast operation for N consecutive days, and the first time period is a time period before the N days; the N days include the current time, and N is an integer greater than 1;

[0006] Extracting features of the operation information of the object to be identified within a second time period to obtain a second operation feature corresponding to the object to be identified; the second time period is a time period after the first time period;

[0007] Based on the first operation feature and the second operation feature, predicting the live broadcast operation of the to-be-identified object to obtain live broadcast prediction information;

[0008] Based on the live broadcast prediction information, a target object for continuing to perform the live broadcast operation is determined from the objects to be identified.

[0009] As an optional embodiment, the first operation feature includes a live broadcast operation feature, and the live broadcast operation of the object to be identified is predicted based on the first operation feature and the second operation feature to obtain the live broadcast prediction information including:

[0010] Based on the weight information corresponding to the live broadcast operation feature, weighted processing is performed on the live broadcast operation feature to obtain a live broadcast operation time sequence feature;

[0011] Determine, according to the time sequence information corresponding to the first time period, a change trend of the first operation feature within the first time period to obtain live broadcast operation trend information;

[0012] Determining, according to the time series information corresponding to the second time period, a change trend of the second operation feature within the second time period to obtain content publishing operation trend information;

[0013] Based on the first operation feature, the second operation feature, the live broadcast operation timing feature, the live broadcast operation trend information and the content publishing operation trend information, the live broadcast operation of the object to be identified is predicted to obtain live broadcast prediction information.

[0014] As an optional embodiment, the weighted processing of the live broadcast operation feature based on the weight information corresponding to the live broadcast operation feature to obtain the live broadcast operation timing feature includes:

[0015] Determine the weight information corresponding to the live broadcast operation feature based on the distance between the time node corresponding to the live broadcast operation feature in the first time period and the interruption time node; the interruption time node is the time node dividing the first time period and the second time period; the weight information is information that increases as the distance decreases;

[0016] Based on the weight information, the live broadcast operation characteristics are weighted to obtain the live broadcast operation timing characteristics.

[0017] As an optional embodiment, the first operation feature includes a live broadcast operation feature, a first content publishing operation feature, and a first interactive operation feature, and the feature extraction of the operation information of the object to be identified within the first time period to obtain the first operation feature corresponding to the object to be identified includes:

[0018] Extracting features of operation information related to the live broadcast operation of the object to be identified within the first time period to obtain the live broadcast operation features;

[0019] Extracting features of operation information related to content publishing operations of the to-be-identified object within the first time period to obtain the first content publishing operation features;

[0020] Feature extraction is performed on the interactive operation information performed by the object to be identified within the first time period to obtain the first interactive operation feature.

[0021] As an optional embodiment, the live broadcast operation feature includes a live broadcast feature corresponding to the object to be identified and a live broadcast association feature corresponding to a first associated object, wherein the first associated object is an object having an association relationship with the object to be identified within the first time period;

[0022] The extracting of features of the operation information related to the live broadcast operation of the object to be identified within the first time period to obtain the live broadcast operation features includes:

[0023] Extracting features of live broadcast operation information performed by the to-be-identified object within the first time period to obtain the live broadcast features;

[0024] Feature extraction is performed on first associated operation information performed by the first associated object within the first time period to obtain the live broadcast associated feature, where the first associated operation information is operation information associated with the production content corresponding to the live broadcast operation information.

[0025] As an optional embodiment, the first content publishing operation feature includes a first content publishing feature corresponding to the object to be identified and a first content publishing association feature corresponding to a first associated object, wherein the first associated object is an object having an association relationship with the object to be identified within the first time period;

[0026] The extracting features of the operation information related to the content publishing operation of the to-be-identified object within the first time period to obtain the first content publishing operation features includes:

[0027] Extracting features of content publishing operations performed by the to-be-identified object within the first time period to obtain the first content publishing features;

[0028] Feature extraction is performed on the second associated operation information performed by the first associated object within the first time period to obtain the first content publishing associated feature; the second associated operation feature is the operation information associated with the production content corresponding to the content publishing operation information within the first time period.

[0029] As an optional embodiment, the first interactive operation feature includes a first interactive feature corresponding to the object to be identified and a second interactive feature corresponding to a first interactive object, and the first interactive object is an object that interacts with the object to be identified based on the content produced by the object to be identified within the first time period;

[0030] The extracting features of the interactive operation information performed by the to-be-identified object within the first time period to obtain the first interactive operation features includes:

[0031] Extracting features of interactive operation information performed by the object to be identified within the first time period to obtain the first interactive features;

[0032] Feature extraction is performed on the interactive operation information performed by the first interactive object in the first time period to obtain the second interactive feature.

[0033] As an optional embodiment, the second operation feature includes a second content publishing operation feature and a second interactive operation feature, and the feature extraction of the operation information of the object to be identified within the second time period to obtain the second operation feature corresponding to the object to be identified includes:

[0034] Extracting features of the content publishing operation information performed by the to-be-identified object within the second time period to obtain the second content publishing operation features;

[0035] Feature extraction is performed on the interactive operation information performed by the object to be identified within the second time period to obtain the second interactive operation feature.

[0036] As an optional embodiment, the second content publishing operation feature includes a second content publishing feature corresponding to the object to be identified and a content publishing association feature corresponding to a second associated object, and the second associated object is an object having an association relationship with the object to be identified within the second time period;

[0037] The extracting features of the operation information related to the content publishing operation of the to-be-identified object within the second time period to obtain the second content publishing operation features includes:

[0038] Extracting features of the content publishing operation information performed by the to-be-identified object within the second time period to obtain the second content publishing features;

[0039] Feature extraction is performed on the third associated operation information performed by the second associated object within the second time period to obtain the second content publishing associated feature; the third associated operation feature is the operation information associated with the production content corresponding to the content publishing operation information within the second time period.

[0040] As an optional embodiment, the second interactive operation feature includes a third interactive feature corresponding to the object to be identified and a fourth interactive feature corresponding to the second interactive object, and the second interactive object is an object that interacts with the object to be identified based on the content produced by the object to be identified within the second time period;

[0041] The extracting features of the interactive operation information performed by the to-be-identified object within the second time period to obtain the second interactive operation features includes:

[0042] Extracting features of interactive operation information performed by the object to be identified within the second time period to obtain the third interactive feature;

[0043] Feature extraction is performed on the interactive operation information performed by the second interactive object in the second time period to obtain the fourth interactive feature.

[0044] As an optional embodiment, the method further includes:

[0045] Extracting features of interactive operation information between the object to be identified and a third interactive object to obtain a third interactive operation feature corresponding to the object to be identified; the third interactive object is an object that interacts with the object to be identified based on an object association relationship corresponding to the object to be identified;

[0046] The predicting the live broadcast operation of the to-be-identified object based on the first operation feature and the second operation feature to obtain the live broadcast prediction information includes:

[0047] Based on the first operation feature, the second operation feature and the third interactive operation feature, the live broadcast operation of the object to be identified is predicted to obtain live broadcast prediction information.

[0048] According to a second aspect of an embodiment of the present disclosure, there is provided an object recognition device, the device comprising:

[0049] A first feature extraction module is configured to perform feature extraction on operation information of an object to be identified within a first time period to obtain a first operation feature corresponding to the object to be identified; the object to be identified is an object that has not performed a live broadcast operation for N consecutive days, and the first time period is a time period before the N days; the N days include the current time, and N is an integer greater than 1;

[0050] A second feature extraction module is configured to perform feature extraction on the operation information of the object to be identified within a second time period to obtain a second operation feature corresponding to the object to be identified; the second time period is a time period after the first time period;

[0051] A live broadcast prediction module is configured to predict the live broadcast operation of the to-be-identified object based on the first operation feature and the second operation feature to obtain live broadcast prediction information;

[0052] The target object determination module is configured to determine the target object for continuing to perform the live broadcast operation from the objects to be identified based on the live broadcast prediction information.

[0053] As an optional embodiment, the first operation feature includes a live broadcast operation feature, and the live broadcast prediction module includes:

[0054] A weighted processing unit, configured to perform weighted processing on the live broadcast operation feature based on the weight information corresponding to the live broadcast operation feature to obtain a live broadcast operation timing feature;

[0055] a live broadcast operation trend determination unit, configured to determine a change trend of the first operation feature within the first time period according to the time sequence information corresponding to the first time period, and obtain live broadcast operation trend information;

[0056] a content publishing operation trend determining unit, configured to determine a change trend of the second operation feature within the second time period according to the time sequence information corresponding to the second time period, and obtain content publishing operation trend information;

[0057] The live broadcast operation prediction unit is configured to predict the live broadcast operation of the object to be identified based on the first operation feature, the second operation feature, the live broadcast operation timing feature, the live broadcast operation trend information and the content publishing operation trend information to obtain live broadcast prediction information.

[0058] As an optional embodiment, the weighted processing unit includes:

[0059] The weight information determining unit is configured to determine the weight information corresponding to the live broadcast operation feature based on the distance between the time node corresponding to the live broadcast operation feature in the first time period and the interruption time node; the interruption time node is the time node dividing the first time period and the second time period; the weight information is information that increases as the distance decreases;

[0060] The timing feature determination unit is configured to perform weighted processing on the live broadcast operation feature based on the weight information to obtain the live broadcast operation timing feature.

[0061] As an optional embodiment, the first operation feature includes a live broadcast operation feature, a first content publishing operation feature, and a first interactive operation feature, and the first feature extraction module includes:

[0062] A live broadcast operation feature extraction unit is configured to perform feature extraction on the operation information related to the live broadcast operation of the object to be identified within the first time period to obtain the live broadcast operation feature;

[0063] A first content publishing operation feature extraction unit is configured to perform feature extraction on operation information related to the content publishing operation of the to-be-identified object within the first time period to obtain the first content publishing operation feature;

[0064] The first interactive operation feature extraction unit is configured to extract features of the interactive operation information performed by the to-be-identified object within the first time period to obtain the first interactive operation feature.

[0065] As an optional embodiment, the live broadcast operation feature includes a live broadcast feature corresponding to the object to be identified and a live broadcast association feature corresponding to a first associated object, wherein the first associated object is an object having an association relationship with the object to be identified within the first time period;

[0066] The live broadcast operation feature extraction unit comprises:

[0067] A live broadcast feature extraction unit is configured to extract features of live broadcast operation information performed by the object to be identified in the first time period to obtain the live broadcast feature;

[0068] The live broadcast associated feature extraction unit is configured to perform feature extraction on the first associated operation information performed by the first associated object within the first time period to obtain the live broadcast associated feature, where the first associated operation information is operation information associated with the production content corresponding to the live broadcast operation information.

[0069] As an optional embodiment, the first content publishing operation feature includes a first content publishing feature corresponding to the object to be identified and a first content publishing association feature corresponding to a first associated object, wherein the first associated object is an object having an association relationship with the object to be identified within the first time period;

[0070] The first content publishing operation feature extraction unit includes:

[0071] A first content publishing feature extraction unit is configured to extract features of content publishing operation information performed by the to-be-identified object within the first time period to obtain the first content publishing feature;

[0072] The first content publishing associated feature extraction unit is configured to perform feature extraction on the second associated operation information performed by the first associated object within the first time period to obtain the first content publishing associated feature; the second associated operation feature is the operation information associated with the production content corresponding to the content publishing operation information within the first time period.

[0073] As an optional embodiment, the first interactive operation feature includes a first interactive feature corresponding to the object to be identified and a second interactive feature corresponding to a first interactive object, and the first interactive object is an object that interacts with the object to be identified based on the content produced by the object to be identified within the first time period;

[0074] The first interactive operation feature extraction unit includes:

[0075] A first interaction feature extraction unit is configured to extract features of interaction operation information performed by the object to be identified within the first time period to obtain the first interaction feature;

[0076] The second interaction feature extraction unit is configured to perform feature extraction on the interaction operation information performed by the first interaction object in the first time period to obtain the second interaction feature.

[0077] As an optional embodiment, the second operation feature includes a second content publishing operation feature and a second interactive operation feature, and the second feature extraction module includes:

[0078] A second content publishing operation feature extraction unit is configured to extract features of the content publishing operation information performed by the to-be-identified object within the second time period to obtain the second content publishing operation feature;

[0079] The second interactive operation feature extraction unit is configured to extract features from the interactive operation information performed by the to-be-identified object within the second time period to obtain the second interactive operation feature.

[0080] As an optional embodiment, the second content publishing operation feature includes a second content publishing feature corresponding to the object to be identified and a content publishing association feature corresponding to a second associated object, and the second associated object is an object having an association relationship with the object to be identified within the second time period;

[0081] The second content publishing operation feature extraction unit includes:

[0082] A second content publishing feature extraction unit is configured to extract features of content publishing operation information performed by the to-be-identified object within the second time period to obtain the second content publishing feature;

[0083] The second content release associated feature extraction unit is configured to perform feature extraction on the third associated operation information performed by the second associated object within the second time period to obtain the second content release associated feature; the third associated operation feature is the operation information associated with the production content corresponding to the content release operation information within the second time period.

[0084] As an optional embodiment, the second interactive operation feature includes a third interactive feature corresponding to the object to be identified and a fourth interactive feature corresponding to the second interactive object, and the second interactive object is an object that interacts with the object to be identified based on the content produced by the object to be identified within the second time period;

[0085] The second interactive operation feature extraction unit includes:

[0086] A third interaction feature extraction unit is configured to extract features of the interaction operation information performed by the object to be identified within the second time period to obtain the third interaction feature;

[0087] The fourth interaction feature extraction unit is configured to perform feature extraction on the interaction operation information performed by the second interaction object in the second time period to obtain the fourth interaction feature.

[0088] As an optional embodiment, the device further includes:

[0089] A third interactive operation feature extraction module is configured to perform feature extraction on the interactive operation information between the object to be identified and a third interactive object to obtain a third interactive operation feature corresponding to the object to be identified; the third interactive object is an object that interacts with the object to be identified based on the object association relationship corresponding to the object to be identified;

[0090] The live broadcast prediction module includes:

[0091] The live broadcast prediction unit is configured to predict the live broadcast operation of the object to be identified based on the first operation feature, the second operation feature and the third interactive operation feature to obtain live broadcast prediction information.

[0092] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, the electronic device comprising:

[0093] processor;

[0094] a memory for storing instructions executable by the processor;

[0095] The processor is configured to execute the instructions to implement the object recognition method as described above.

[0096] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the object recognition method as described above.

[0097] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the object recognition method described above is implemented.

[0098] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0099] Feature extraction is performed on the operation information of the object to be identified within the first time period and the operation information within the second time period to obtain the first operation feature and the second operation feature corresponding to the object to be identified. The object to be identified is an object that has not broadcast live for N consecutive days, and the first time period is the time period before the N days; N days include the current time, and N is an integer greater than 1. Based on the first operation feature and the second operation feature, it is determined whether the object to be identified will continue to perform the live broadcast operation. This method integrates the features corresponding to the live broadcast information before the object to be identified interrupts the live broadcast, and the features corresponding to the activity level of the object to be identified on the target application after the object to be identified interrupts the live broadcast, and integrates multi-dimensional features to identify potential target objects, thereby improving the accuracy and effectiveness of object recognition, and also improving the interactive experience between the object to be identified and the interactive object, forming positive interactive feedback.

[0100] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0101] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0102] Figure 1 The figure is a schematic diagram of an application scenario of an object recognition method according to an exemplary embodiment.

[0103] Figure 2 The figure is a flow chart of an object recognition method according to an exemplary embodiment.

[0104] Figure 3 The present invention is a flowchart of obtaining a first operation feature in an object recognition method according to an exemplary embodiment.

[0105] Figure 4 The present invention is a flowchart of obtaining live broadcast operation features in an object recognition method according to an exemplary embodiment.

[0106] Figure 5 The present invention is a flowchart of obtaining a first content publishing operation feature in an object recognition method according to an exemplary embodiment.

[0107] Figure 6 The present invention is a flowchart of obtaining a first interactive operation feature in an object recognition method according to an exemplary embodiment.

[0108] Figure 7 The present invention is a flowchart of obtaining a second operation feature in an object recognition method according to an exemplary embodiment.

[0109] Figure 8 The present invention is a flowchart of obtaining a second content publishing operation feature in an object recognition method according to an exemplary embodiment.

[0110] Fig. 9 The present invention is a flowchart of obtaining a second interactive operation feature in an object recognition method according to an exemplary embodiment.

[0111] Fig.10 The present invention is a flowchart of determining live broadcast operation trend information and content publishing operation trend information in an object recognition method according to an exemplary embodiment.

[0112] Fig.11 It is a schematic diagram of a change trend of the live broadcast operation of the entire object to be identified before the live broadcast operation is interrupted in an object recognition method according to an exemplary embodiment.

[0113] Fig.12 The diagram is a schematic diagram showing the division of a first time period, a second time period and a preset time period in an object recognition method according to an exemplary embodiment.

[0114] Fig.13 The present invention is a flowchart of weighted processing of live broadcast features in an object recognition method according to an exemplary embodiment.

[0115] Fig.14 It is a schematic diagram of the reflow rate of a preset object corresponding to different interruption duration information in an object recognition method according to an exemplary embodiment.

[0116] Fig.15 It is a flowchart of performing live broadcast prediction based on a first operation feature, a second operation feature, and a third interactive operation feature in an object recognition method according to an exemplary embodiment.

[0117] Fig.16 The figure is a block diagram of an object recognition device according to an exemplary embodiment.

[0118] Fig.17 FIG. 1 is a block diagram of an electronic device according to an exemplary embodiment. (General structure of a server) DETAILED DESCRIPTION

[0119] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0120] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0121] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0122] Figure 1 is a schematic diagram of an application scenario of an object recognition method according to an exemplary embodiment. Figure 1 As shown, the application scenario includes a client 110 and a server 120. The server 120 obtains the operation information within the first time period and the operation information within the second time period corresponding to the objects to be identified of each of the multiple clients 110. The server 120 performs feature extraction on the operation information of the object to be identified within the first time period to obtain the first operation feature corresponding to the object to be identified, and performs feature extraction on the operation information of the object to be identified within the second time period to obtain the second operation feature corresponding to the object to be identified. Based on the first operation feature and the second operation feature, the server 120 predicts the live broadcast operation of the object to be identified and obtains live broadcast prediction information. Based on the live broadcast prediction information, the target object that continues to perform the live broadcast operation is determined from the object to be identified. The server 120 sends object recall information to the client 110 corresponding to the target object.

[0123] In the embodiments of the present disclosure, the client 110 includes physical devices such as smart phones, desktop computers, tablet computers, laptops, digital assistants, smart wearable devices, etc., and may also include software running in the physical devices, such as applications, etc. The operating system running on the physical device in the embodiments of the present application may include but is not limited to Android system, IOS system, Linux, Unix, Windows, etc.

[0124] In the embodiment of the present disclosure, the server 120 may include an independently operated server, or a distributed server, or a server cluster composed of multiple servers. The server 120 may include a network communication unit, a processor, a memory, and the like.

[0125] Figure 2 is a flow chart of an object recognition method according to an exemplary embodiment. Figure 2 As shown, the method is used in a server and includes the following steps.

[0126] S210. Extracting features from the operation information of the object to be identified within the first time period to obtain a first operation feature corresponding to the object to be identified; the object to be identified is an object that has not performed a live broadcast operation for N consecutive days, and the first time period is the time period before N days; N days includes the current time, and N is an integer greater than 1;

[0127] As an optional embodiment, the first operation feature is a feature extracted based on the operation information of the object to be identified before the broadcast is stopped. The first operation feature may include a live broadcast operation feature, a first content publishing operation feature, and a first interactive operation feature. The first operation feature is used to feedback the live broadcast-related activity level and feedback information of the object to be identified before the live broadcast operation is interrupted.

[0128] As an alternative embodiment, see Figure 3 The first operation feature includes a live broadcast operation feature, a first content publishing operation feature, and a first interactive operation feature. The operation information of the object to be identified in the first time period is subjected to feature extraction, and the first operation feature corresponding to the object to be identified includes:

[0129] S310. Extracting features of the operation information related to the live broadcast operation of the identified object in the first time period to obtain live broadcast operation features;

[0130] S320. Extracting features of the operation information related to the content publishing operation of the identified object within the first time period to obtain a first content publishing operation feature;

[0131] S330. Extract features from the interactive operation information performed by the object to be identified within the first time period to obtain a first interactive operation feature.

[0132] As an optional embodiment, feature extraction is performed on the operation information related to the live broadcast, the operation information related to the content publishing operation, and the interactive operation information in the first time period, and a live broadcast operation feature, a first content publishing operation feature, and a first interactive operation feature can be obtained. The live broadcast operation feature can represent the live broadcast state of the object to be identified before the live broadcast operation is interrupted, and can include features actively executed by the object to be identified and features actively executed by the first associated object. The first content publishing operation feature can represent the content publishing state of the object to be identified before the live broadcast operation is interrupted, and the content publishing state may be related to the live broadcast or unrelated to the live broadcast, and can also include features actively executed by the object to be identified and features actively executed by the first associated object. The first interactive operation feature can represent the interactive state between the object to be identified and the first interactive object before the live broadcast operation is interrupted, and can include features actively executed by the object to be identified and features actively executed by the first interactive object.

[0133] Through various characteristic information within the first time period, the activity level of the object to be identified before the live broadcast operation is interrupted and the feedback information of the first associated object on the production content of the object to be identified can be determined. The higher the activity level before the live broadcast operation is interrupted and the more positive feedback there are, the higher the possibility that the object to be identified will continue the live broadcast; the lower the activity level before the live broadcast operation is interrupted and the fewer positive feedback there are, the lower the possibility that the object to be identified will continue the live broadcast.

[0134] By acquiring active and passive features in three different dimensions of live broadcast operation, content publishing operation and interactive operation within the first time period, the activity level and feedback information of the object to be identified before the live broadcast operation is interrupted can be determined through multiple dimensions, thereby determining the live broadcast state of the object to be identified before the live broadcast operation is interrupted, improving the comprehensiveness of the first operation feature, facilitating the subsequent steps of executing the live broadcast operation prediction, and improving the accuracy of the live broadcast operation prediction.

[0135] As an alternative embodiment, see Figure 4 The live broadcast operation feature includes a live broadcast feature corresponding to the object to be identified and a live broadcast association feature corresponding to the first associated object, where the first associated object is an object that has an association relationship with the object to be identified within a first time period;

[0136] Feature extraction is performed on the operation information related to the live broadcast operation of the object to be identified in the first time period, and the live broadcast operation features obtained include:

[0137] S410. Extracting features from the live broadcast operation information performed by the identified object within the first time period to obtain live broadcast features;

[0138] S420. Perform feature extraction on the first associated operation information executed by the first associated object within the first time period to obtain live broadcast associated features, where the first associated operation information is operation information associated with the production content corresponding to the live broadcast operation information.

[0139] As an optional embodiment, in the first operation feature, the live broadcast feature is a feature obtained based on the operation information actively performed by the object to be identified. The live broadcast feature can be a feature obtained based on the live broadcast operation information performed by the object to be identified, such as live broadcast frequency, live broadcast duration, etc.

[0140] Based on the live broadcast characteristics, the live broadcast activity of the object to be identified before the live broadcast is interrupted can be directly obtained. Among them, the live broadcast activity can reflect the possibility of the object to be identified to be live broadcast. The greater the activity, the greater the possibility of the object to be identified to be live broadcast, and there is a higher possibility to continue the live broadcast after the live broadcast is interrupted. The smaller the activity, the lower the possibility of the object to be identified to be live broadcast, and there is a lower possibility to continue the live broadcast after the live broadcast is interrupted.

[0141] In the first operation feature, the live broadcast associated feature is a feature obtained based on the operation information passively performed by the object to be identified, that is, a feature obtained based on the operation information actively performed by the first associated object. The first associated object may be an object that has a social relationship with the object to be identified, such as a fan or friend of the anchor. The first associated operation information actively performed by the first associated object may include live broadcast connection, live broadcast slicing, secondary creation, reprinting, etc. Secondary creation may be a cover, remake, editing, drawing, etc. The live broadcast associated feature may be a feature obtained based on the operation information related to the live broadcast operation information in the first associated operation information performed by the first associated object. The live broadcast associated feature may include the frequency of friends' live broadcast connections, the frequency of fans' live broadcast slicing, the number of live broadcast reprints, etc.

[0142] Based on the live broadcast association features, the degree to which the associated object participated in the live broadcast before the live broadcast was interrupted can be obtained. Among them, the degree to which the associated object participated in the live broadcast can reflect the possibility of the associated object for the live broadcast, and can also reflect the feedback information of the object to be identified during the live broadcast. The higher the degree of the associated object's participation in the live broadcast, the greater the possibility of the associated object for the live broadcast, and the more positive feedback the object to be identified has during the live broadcast, the higher the possibility of the object to be identified continuing the live broadcast after the live broadcast is interrupted. The lower the degree of the associated object's participation in the live broadcast, the lower the possibility of the associated object for the live broadcast, and the more negative feedback the object to be identified has during the live broadcast, the lower the possibility of the object to be identified continuing the live broadcast after the live broadcast is interrupted.

[0143] By acquiring the live broadcast features actively executed by the object to be identified and the live broadcast associated features executed by the first associated object, the relevant circumstances of the live broadcast of the object to be identified can be determined from both active and passive aspects, thereby improving the comprehensiveness of the live broadcast operation features, facilitating the subsequent steps of executing the live broadcast operation prediction, and improving the accuracy of the live broadcast operation prediction.

[0144] As an alternative embodiment, see Figure 5 , the first content publishing operation feature includes a first content publishing feature corresponding to the object to be identified and a first content publishing association feature corresponding to the first associated object, the first associated object being an object having an association relationship with the object to be identified within the first time period;

[0145] Feature extraction is performed on operation information related to the content publishing operation of the object to be identified within the first time period, and the first content publishing operation feature obtained includes:

[0146] S510. Extracting features from the content publishing operation information performed by the identified object within the first time period to obtain a first content publishing feature;

[0147] S520. Perform feature extraction on the second associated operation information performed by the first associated object within the first time period to obtain a first content publishing associated feature; the second associated operation feature is operation information associated with the production content corresponding to the content publishing operation information within the first time period.

[0148] As an optional embodiment, in the first operation feature, the first content publishing feature is a feature obtained based on the operation information actively performed by the object to be identified within the first time period. The first content publishing feature can be a feature obtained based on the content publishing operation information performed by the object to be identified, such as content publishing frequency, content publishing duration, etc. Based on the first content publishing feature, the degree of activity of the object to be identified in executing content publishing before the live broadcast is interrupted can be directly obtained. Among them, the degree of activity in executing content publishing can reflect the possibility of the object to be identified executing content publishing. The greater the degree of activity, the greater the possibility of the object to be identified executing content publishing. If the content publishing is related to the live broadcast, it means that the object to be identified is more active in creating content related to the live broadcast, which reflects that the object to be identified is more active in live broadcasting, so the possibility of continuing the live broadcast after the live broadcast is interrupted is higher. If the content publishing is not related to the live broadcast, it means that the object to be identified is more active in content creation and less active in live broadcasting, so the possibility of continuing the live broadcast after the live broadcast is interrupted is lower.

[0149] In the first operation feature, the first content release associated feature is a feature obtained based on the operation information passively performed by the object to be identified, that is, a feature obtained based on the operation information actively performed by the first associated object. The first associated object may be an object having a social relationship with the object to be identified, such as a fan or friend of the anchor. The second associated operation information actively performed by the first associated object may include secondary creation, reprinting, etc. Secondary creation may be a cover, remake, editing, drawing, etc. The first content release associated feature may be a feature obtained based on the operation information related to the content release operation information in the second associated operation information performed by the first associated object. The first content release associated feature may include the number of secondary creations, the popularity of secondary creations, the number of content reprints, etc.

[0150] Based on the first content publishing association feature, the degree to which the first associated object participated in executing content publishing before the live broadcast operation was interrupted can be obtained. The degree to which the first associated object participated in executing content publishing can reflect the possibility of the associated object creating content related to the live broadcast, and can also reflect the feedback information when the object to be identified executes content creation related to the live broadcast.

[0151] If the content release is related to live broadcast, the higher the degree of participation of the first associated object in executing content release, the greater the possibility of the first associated object creating content related to live broadcast, the more positive feedback the object to be identified has when executing content release, and the higher the possibility of the object to be identified continuing live broadcast after interrupting the live broadcast operation. The lower the degree of participation of the first associated object in executing content release, the lower the possibility of the first associated object creating content related to live broadcast, the more negative feedback the object to be identified has when executing content release, and the lower the possibility of the object to be identified continuing live broadcast after interrupting the live broadcast operation.

[0152] If the content release is not related to live broadcast, the higher the degree of participation of the first associated object in executing content release, the greater the possibility of the first associated object in content creation, the more positive feedback the object to be identified has when executing content release, and the lower the possibility of the object to be identified continuing live broadcast after interrupting the live broadcast operation. The lower the degree of participation of the first associated object in executing content release, the lower the possibility of the first associated object in content creation, the more negative feedback the object to be identified has when executing content release, and the object to be identified may stop content release or no longer use the target application.

[0153] By obtaining the first content publishing feature actively executed by the object to be identified and the first content publishing associated feature executed by the first associated object, the content publishing information of the object to be identified within the first time period can be determined from both active and passive aspects, among which the content release related to the live broadcast content can be used as a reference for the live broadcast status, thereby improving the comprehensiveness of the first content publishing operation feature, facilitating the subsequent steps of executing the live broadcast operation prediction, and improving the accuracy of the live broadcast operation prediction.

[0154] As an alternative embodiment, see Figure 6 The first interactive operation feature includes a first interactive feature corresponding to the object to be identified and a second interactive feature corresponding to the first interactive object, and the first interactive object is an object that interacts with the object to be identified based on the production content of the object to be identified within the first time period;

[0155] Feature extraction is performed on the interactive operation information performed by the object to be identified within the first time period, and the first interactive operation feature obtained includes:

[0156] S610. Extracting features of interactive operation information performed by the object to be identified within the first time period to obtain a first interactive feature;

[0157] S620. Perform feature extraction on the interactive operation information performed by the first interactive object within the first time period to obtain a second interactive feature.

[0158] As an optional embodiment, the first interaction feature is obtained based on the operation information actively performed by the object to be identified. The first interaction feature may include the viewing time, virtual resource information, like information, comment information, etc. of the object to be identified. The first interaction feature may reflect the activity level of the object to be identified in the target application. In the case where the activity level corresponding to the first interaction feature is higher, it means that the frequency of the object to be identified using the target application is higher, and the object to be identified is an active user of the target application. In this case, there is a higher possibility that the object to be identified will continue the live broadcast after the live broadcast is interrupted. In the case where the activity level corresponding to the first interaction feature is lower, it means that the frequency of the object to be identified using the target application is lower, and the object to be identified has reduced the use of the target application or does not use the target application. In this case, there is a lower possibility that the object to be identified will continue the live broadcast after the live broadcast is interrupted.

[0159] The second interaction feature is obtained based on the operation information passively performed by the object to be identified, that is, based on the interaction operation information actively performed by the interactive object. The interactive object may include a first interaction object, and the first interaction object is an object that interacts with the object to be identified based on the live content and published content within the first time period. When interacting based on the live content within the first time period, the first interaction object may be an object that interacts with the object to be identified in the same virtual space, and the virtual space may be a live broadcast room. The first interaction operation feature corresponding to the first interaction object may include live viewing time, virtual resource information, live like information, live comment information, live viewing number, etc.

[0160] The second interaction feature can reflect the feedback information of the interactive object on the production content of the object to be identified. In the case where the positive feedback of the live broadcast content in the second interaction feature corresponding to the first interactive object is higher, it means that the interactive object has more positive feedback on the live broadcast content, and therefore, the object to be identified is more likely to continue the live broadcast after interrupting the live broadcast.

[0161] By obtaining the first interaction feature actively executed by the object to be identified and the second interaction feature executed by the first interaction object, the interaction information of the object to be identified within the first time period can be determined from both active and passive aspects, wherein the interaction information related to the live broadcast content can be used as a reference for the live broadcast status, thereby improving the comprehensiveness of the first content publishing operation feature, facilitating the subsequent steps of executing the live broadcast operation prediction, and improving the accuracy of the live broadcast operation prediction.

[0162] S220. Extracting features from the operation information of the object to be identified in the second time period to obtain a second operation feature corresponding to the object to be identified; the second time period is a time period after the first time period;

[0163] As an optional embodiment, the second operation feature is a feature extracted based on the operation information after the object to be identified stops broadcasting. The second operation feature does not include the live broadcast operation feature, and the second operation feature may include the first content publishing operation feature and the second interactive operation feature. The second operation feature is used to feedback the live broadcast-related activity level and feedback information of the object to be identified after the broadcasting stops.

[0164] As an alternative embodiment, see Figure 7 The second operation feature includes a second content publishing operation feature and a second interactive operation feature. The operation information of the object to be identified in the second time period is subjected to feature extraction, and the second operation feature corresponding to the object to be identified is obtained, including:

[0165] S710. Extracting features of the content publishing operation information performed by the identified object in the second time period to obtain a second content publishing operation feature;

[0166] S720. Perform feature extraction on the interactive operation information performed by the object to be identified within the second time period to obtain a second interactive operation feature.

[0167] As an optional embodiment, feature extraction is performed on the operation information and interactive operation information related to the content publishing operation in the second time period to obtain a second content publishing operation feature and a second interactive operation feature. The second content publishing operation feature may represent the content publishing status of the object to be identified after the broadcast is stopped. Since the object to be identified has been stopped, the content publishing status may be unrelated to the live broadcast, and may include features actively executed by the object to be identified and features actively executed by the second associated object. The second interactive operation feature may represent the interactive status between the object to be identified and the second interactive object after the broadcast is stopped, and may include features actively executed by the object to be identified and features actively executed by the second interactive object.

[0168] Through various characteristic information within the second time period, the activity level of the object to be identified after the suspension and the feedback information of the second associated object on the content produced by the object to be identified can be determined. The higher the activity level after the suspension and the more positive feedback there are, the lower the possibility that the object to be identified will continue to broadcast live. The lower the activity level before the suspension and the fewer positive feedback there are, the object to be identified may stop publishing content or no longer use the target application.

[0169] By acquiring active and passive features in two different dimensions of content publishing operation and interactive operation in the second time period, the activity level and feedback information of the object to be identified after the suspension can be determined through multiple dimensions, thereby determining the content publishing status or application usage status of the object to be identified after the suspension, thereby improving the comprehensiveness of the second operation feature, facilitating the subsequent steps of executing the live broadcast operation prediction, and improving the accuracy of the live broadcast operation prediction.

[0170] As an alternative embodiment, see Figure 8 The second content publishing operation feature includes a second content publishing feature corresponding to the object to be identified and a content publishing association feature corresponding to the second associated object, where the second associated object is an object that has an association relationship with the object to be identified within the second time period;

[0171] Feature extraction is performed on the operation information related to the content publishing operation of the object to be identified within the second time period, and the second content publishing operation feature obtained includes:

[0172] S810. Extracting features of the content publishing operation information performed by the identified object within the second time period to obtain a second content publishing feature;

[0173] S820. Perform feature extraction on the third associated operation information executed by the second associated object within the second time period to obtain a second content publishing associated feature; the third associated operation feature is operation information associated with the production content corresponding to the content publishing operation information within the second time period.

[0174] As an optional embodiment, in the second operation feature, the second content publishing feature is a feature obtained based on the operation information actively performed by the object to be identified in the second time period. The second content publishing feature may be a feature obtained based on the content publishing operation information performed by the object to be identified, such as content publishing frequency, content publishing duration, etc. Based on the second content publishing feature, the degree of activity of the object to be identified in executing content publishing after the live broadcast is interrupted can be directly obtained. Among them, the degree of activity in executing content publishing can reflect the possibility of the object to be identified executing content publishing. The greater the degree of activity, the greater the possibility that the object to be identified executes content publishing, and there is a lower possibility of continuing the live broadcast after the live broadcast is interrupted. If the degree of activity is lower than the preset activity threshold, the object to be identified may also stop publishing content or no longer use the target application.

[0175] In the first operation feature, the second content release association feature is a feature obtained based on the operation information passively performed by the object to be identified, that is, a feature obtained based on the operation information actively performed by the second associated object. The second associated object may be an object having a social relationship with the object to be identified, such as the anchor's fans, friends, etc. The second associated operation information actively performed by the second associated object may include secondary creation, reprinting, etc. Secondary creation may be a cover, remake, editing, drawing, etc. The second content release association feature may be a feature obtained based on the operation information related to the content release operation information in the second associated operation information performed by the first associated object. The first content release association feature may include the number of secondary creations, the popularity of secondary creations, the number of content reprints, etc.

[0176] Based on the second content publishing association feature, the degree to which the associated object participates in executing content publishing after the live broadcast operation is interrupted can be obtained. Among them, the degree to which the second associated object participates in executing content publishing can reflect the possibility of the associated object for content publishing, and can also reflect the feedback information when the object to be identified executes content publishing. The higher the degree to which the associated object participates in executing content publishing, the greater the possibility of the associated object for content publishing, and the more positive feedback the object to be identified receives when executing content publishing, the lower the possibility that the object to be identified will continue the live broadcast after the live broadcast operation is interrupted. Similarly, if the degree to which the second associated object participates in executing content publishing reflected by the second content publishing association feature is lower than the preset participation level threshold, the object to be identified may also stop publishing content or no longer use the target application.

[0177] By acquiring the second content publishing feature actively executed by the object to be identified and the second content publishing associated feature executed by the second associated object, the content publishing information of the object to be identified in the second time period can be obtained from both active and passive aspects, thereby determining the content publishing status of the object to be identified after the broadcast is stopped, thereby improving the comprehensiveness of the second content publishing operation feature, facilitating the subsequent execution of the live broadcast operation prediction step, and improving the accuracy of the live broadcast operation prediction.

[0178] As an alternative embodiment, see Fig. 9 , the second interactive operation feature includes a third interactive feature corresponding to the object to be identified and a fourth interactive feature corresponding to the second interactive object, the second interactive object being an object that interacts with the object to be identified based on the production content of the object to be identified within the second time period;

[0179] Feature extraction is performed on the interactive operation information performed by the object to be identified in the second time period to obtain the second interactive operation feature, including:

[0180] S910. Extracting features from the interactive operation information performed by the object to be identified within the second time period to obtain a third interactive feature;

[0181] S920. Perform feature extraction on the interactive operation information performed by the second interactive object in the second time period to obtain a fourth interactive feature.

[0182] As an optional embodiment, the third interaction feature is obtained based on the operation information actively performed by the object to be identified. The third interaction feature may include the viewing time, virtual resource information, like information, comment information, etc. of the object to be identified in the second time period. The third interaction feature can reflect the activity level of the object to be identified in the target application. In the case where the activity level corresponding to the third interaction feature is higher, it means that the frequency of the object to be identified using the target application is higher, and the object to be identified is still an active user of the target application. In this case, there is a higher possibility that the object to be identified will continue the live broadcast after the live broadcast is interrupted. In the case where the activity level corresponding to the third interaction feature is lower, it means that the frequency of the object to be identified using the target application is lower, and the object to be identified has reduced the use of the target application or does not use the target application. In this case, there is a lower possibility that the object to be identified will continue the live broadcast after the live broadcast is interrupted.

[0183] The fourth interactive feature is obtained based on the operation information passively performed by the object to be identified, that is, based on the interactive operation information actively performed by the interactive object. The interactive object may be a second interactive object, and the second interactive object is an object that interacts with the object to be identified by publishing content within a second time period. The second interactive operation feature corresponding to the second interactive object may include video completion rate, video like information, video comment information, video viewing number, etc.

[0184] The fourth interaction feature may reflect the feedback information of the interactive object on the content produced by the object to be identified. In the case where the positive feedback information in the fourth interaction feature corresponding to the second interactive object is higher, it means that the second interactive object has more positive feedback on the content published in the second time period. Therefore, the object to be identified is less likely to continue the live broadcast after the live broadcast is interrupted.

[0185] By obtaining the first interaction feature actively executed by the object to be identified and the second interaction feature executed by the first interaction object, the interaction information of the object to be identified within the first time period can be determined from both active and passive aspects, wherein the interaction information related to the live broadcast content can be used as a reference for the live broadcast status, thereby improving the comprehensiveness of the first content publishing operation feature, facilitating the subsequent steps of executing the live broadcast operation prediction, and improving the accuracy of the live broadcast operation prediction.

[0186] S230. Based on the first operation feature and the second operation feature, predict the live broadcast operation of the object to be identified to obtain live broadcast prediction information;

[0187] S240. Based on the live broadcast prediction information, determine the target object for continuing to perform the live broadcast operation from the objects to be identified.

[0188] As an optional embodiment, based on the first operation feature, the second operation feature and the social operation feature, the live broadcast operation of the object to be identified is predicted to obtain live broadcast prediction information, and based on the live broadcast prediction information, a target object that continues to broadcast live information within a preset time period is determined from the object to be identified. The preset time period can be a general period in object recall, which can be 7 days.

[0189] As an optional embodiment, the first operation feature and the second operation feature can be input into a pre-trained object recognition model for object recognition processing, and the target object for continuing to broadcast information can be determined from the objects to be recognized. The object recognition model can be obtained by model training using a training data set constructed by the first training operation feature, the second training operation feature, and the social training operation feature. Among them, the first training operation feature is obtained by feature extraction of sample operation information within the first time period corresponding to the sample object, and the second training operation feature is obtained by feature extraction of sample operation information within the second time period corresponding to the sample object. The first training operation feature includes features of live broadcast, content publishing, and interaction, and the second training operation feature includes features of content publishing and interaction.

[0190] The object recognition model can be a gradient boosting algorithm model, such as an XGboost model. The gradient boosting algorithm model can update the model parameters by fitting the residual tree with preset residual gain information in each iteration until the model converges. The parameter gain information of the XGboost model is the structural score before the residual tree split minus the structural score after the residual tree split, and the segmentation point with the largest gain is selected as the optimal segmentation point.

[0191] During the model training process, the training data set can also be divided into a training set, a validation set, and a test set according to a preset ratio. Based on the validation set, the model parameters of the object recognition model obtained after model training can be optimized.

[0192] As an alternative embodiment, see Fig.10 The first operation feature includes a live broadcast operation feature. Based on the first operation feature and the second operation feature, the live broadcast operation of the object to be identified is predicted, and the obtained live broadcast prediction information includes:

[0193] S1010. Based on the weight information corresponding to the live operation characteristics, the live operation characteristics are weighted to obtain the live operation timing characteristics;

[0194] S1020. According to the timing information corresponding to the first time period, the change trend of the first operation feature in the first time period is determined to obtain live operation trend information;

[0195] S1030. According to the timing information corresponding to the second time period, the change trend of the second operation feature in the second time period is determined to obtain content publishing operation trend information;

[0196] S1040. Based on the first operation feature, the second operation feature, the live broadcast operation timing feature, the live broadcast operation trend information and the content publishing operation trend information, the live broadcast operation of the object to be identified is predicted to obtain live broadcast prediction information.

[0197] As an optional embodiment, the timing information corresponding to the first time period may be multiple first sub-time periods, and the timing information corresponding to the second time period may be multiple second sub-time periods. The time interval corresponding to the first sub-time period is the same as the time interval corresponding to the second sub-time period. For example, if the first sub-time period is 7 days, then the second sub-time period is also 7 days.

[0198] The first time period may be a time period in which the trend of the live broadcast information includes a downward trend before the object to be identified interrupts the live broadcast. Fig.11 ,like Fig.11The figure shows the overall trend of the live broadcast operation of the object to be identified before the live broadcast operation is interrupted. The overall trend of the live broadcast operation can be determined by the proportion of the live broadcast object to the objects to be identified. Within 30 days before the live broadcast operation of the object to be identified is interrupted, the proportion information has a significant downward trend. The significant downward trend is the target downward trend, and the first time period can be set to 30 days. The second time period can be the time period from the interruption of the live broadcast of the object to be identified to the time when the object to be identified is recalled.

[0199] See also Fig.12 ,like Fig.12 The diagram shows the division of the first time period, the second time period and the preset time period. The first time period is the 30 days before the live broadcast of the object to be identified is interrupted, the second time period is the time period from the interruption of the live broadcast of the object to be identified to the recall of the object, which can be between 60 days and 90 days, and the preset time period is the time period for predicting the possibility of the object to be identified being recalled, which can be 7 days.

[0200] Based on the live broadcast frequency feature in the live broadcast feature in each first sub-time period, the live broadcast operation trend information can be determined. Based on the content publishing frequency feature in the content publishing feature in each second sub-time period, the content publishing operation trend information can be determined. The live broadcast operation trend information can reflect the change trend of the live broadcast information in the first time period, and the content publishing operation trend information can reflect the change trend of the content publishing information in the second time period.

[0201] When the live broadcast operation trend information indicates that the live broadcast information decreased in the first time period, and when the content publishing operation trend information indicates that the content publishing information increased in the second time period, it means that the object to be identified may be converted from a live broadcast user to a content creation user, and further indicates that the object to be identified has a low possibility of continuing the live broadcast after the live broadcast is interrupted.

[0202] When the live broadcast operation trend information indicates that the live broadcast information has decreased in the first time period, and when the content publishing operation trend information indicates that the content publishing information has decreased in the second time period, it indicates that the object to be identified may be converted from a live broadcast user to a lost user, and further indicates that the object to be identified has a low possibility of continuing the live broadcast after the live broadcast is interrupted.

[0203] When the change trends indicated by the live broadcast operation trend information and the change trends indicated by the content publishing operation trend information are not obvious, it can be determined in combination with other features whether the object to be identified is likely to continue the live broadcast.

[0204] By using live broadcast operation trend information and content publishing operation trend information, the importance of operation features on the content production side in features of each dimension is increased, so that object recognition focuses on the content production capabilities of the object to be identified, thereby improving the success rate of predicting whether the object to be identified will continue to broadcast live and improving the accuracy of object recognition.

[0205] As an alternative embodiment, see Fig.13 Based on the weight information corresponding to the live broadcast operation features, the live broadcast operation features are weighted and the live broadcast operation time series features are obtained, including:

[0206] S1310. Determine the weight information corresponding to the live operation feature based on the distance between the time node corresponding to the live operation feature in the first time period and the interruption time node; the interruption time node is the time node dividing the first time period and the second time period; the weight information is information that increases as the distance decreases;

[0207] S1320. Based on the weight information, weighted processing is performed on the live broadcast operation characteristics to obtain the live broadcast operation timing characteristics.

[0208] As an optional embodiment, for the live broadcast feature within the first time period, the weight information corresponding to the live broadcast feature can be determined based on the distance between the time interval corresponding to the live broadcast feature and the interruption time node. The time interval corresponding to the live broadcast feature can be the time interval corresponding to the first sub-time period, or it can be the time interval corresponding to the third sub-time period. The third sub-time period is obtained after the first time period is divided into intervals. The time interval corresponding to the first sub-time period is different from the time interval corresponding to the third sub-time period. When the distance between the time interval corresponding to the live broadcast feature and the interruption time node is closer, the weight information corresponding to the live broadcast feature is greater. When the distance between the time interval corresponding to the live broadcast feature and the interruption time node is farther, the weight information corresponding to the live broadcast feature is smaller.

[0209] Based on the weight information, the live broadcast characteristics within the first time period are weighted and processed to obtain the live broadcast operation timing characteristics. The live broadcast operation timing characteristics are accumulated and summed to obtain the target content production characteristics corresponding to the live broadcast operation within the first time period. When the target content production characteristics are greater than or equal to the preset feature threshold, it means that the object to be identified has a higher possibility of continuing the live broadcast after the live broadcast is interrupted. When the target content production characteristics are less than the preset feature threshold, it means that the object to be identified has a lower possibility of continuing the live broadcast after the live broadcast is interrupted.

[0210] For example, if the object to be identified performs live broadcasts many times before the interruption time node, the operation duration corresponding to the live broadcast shows a fluctuating trend, and the operation duration in a time interval closest to the interruption time node is higher, then by performing weighted summation on the target feature information in the above-mentioned first time period, a first target content production feature greater than a preset feature threshold can be obtained, indicating that the object to be identified has a higher possibility of continuing the live broadcast.

[0211] If the object to be identified performs live broadcasts many times before the interruption time node, and the operation duration corresponding to each live broadcast is less than the preset duration, then by performing weighted summation on the target feature information within the above-mentioned first time period, a second target content production feature greater than the preset feature threshold but less than the first target content production feature may be obtained, indicating that the object to be identified may continue to broadcast live, but compared with the object to be identified corresponding to the first target content production feature, the object to be identified corresponding to the second target content production feature is less likely to continue to broadcast live.

[0212] The specific formula for calculating the target content production characteristics is as follows:

[0213]

[0214] Among them, S is the target content production feature, t is the time interval corresponding to the live broadcast feature in the first time period, and λ t is the weight information corresponding to the live broadcast feature in the first time period, x t It is the live broadcast feature in the first time period.

[0215] The live broadcast features are weighted to increase their importance in the features of each dimension, so that object recognition focuses on the live broadcast capability of the object to be identified, thereby improving the prediction success rate of the continued live broadcast of the object to be identified and improving the accuracy of object recognition.

[0216] As an optional embodiment, before extracting features of the operation information of the object to be identified within the first time period to obtain the first operation features corresponding to the object to be identified, the method further includes:

[0217] Obtain interruption duration information corresponding to each of the multiple preset objects that interrupt the live broadcast operation;

[0218] The preset objects whose interruption duration information is less than or equal to the preset interruption duration threshold are taken as the objects to be identified.

[0219] As an optional embodiment, the interruption duration information corresponding to each of the multiple preset objects that interrupt the live broadcast operation is obtained, and the interruption duration information is compared with the preset interruption duration threshold to obtain a first comparison result. When the first comparison result indicates that the interruption duration information is less than or equal to the preset interruption duration threshold, it can be determined that the possibility of the preset object continuing the live broadcast is high, and the preset object can be used as the object to be identified. When the first comparison result indicates that the interruption duration information is greater than the preset interruption duration threshold, it can be determined that the possibility of the preset object continuing the live broadcast is low.

[0220] In the case where there are objects to be identified that are difficult to identify, the interruption duration information can also be compared with the preset object recognition threshold to obtain a second comparison result. The object recognition threshold is the duration of the live broadcast interruption of the object to be identified when the change in the probability of successful recognition tends to be stable. The preset objects whose interruption duration information is greater than or equal to the object recognition threshold and less than or equal to the interruption duration threshold can be used as objects to be identified. When the change in the probability of successful recognition tends to be stable, the probability of successful recognition is already less than the preset probability value, and it is difficult to determine whether the object to be identified is the target object.

[0221] See also Fig.14 ,like Fig.14 The above is a schematic diagram of the reflux rate of preset objects corresponding to different interruption duration information. The preset object can be a host object. 60 days after the preset object interrupts the live broadcast operation, the reflux rate is 15.29%. At this time, the reflux rate drops to the inflection point of the curve and then begins to stabilize. 60 days can be used as the object recognition threshold of the preset object. 90 days after the preset object interrupts the live broadcast operation, the reflux rate drops to less than 10%. At this time, the possibility of some preset objects reflux is low, and 90 days can be used as the interruption duration threshold of the preset object.

[0222] Based on the interruption duration information, the objects to be identified are screened, and preset objects with a low possibility of continuing live broadcast are filtered out, and preset objects with a relatively easy possibility of continuing live broadcast can be filtered out, thereby reducing the amount of calculation and improving the efficiency of object recognition.

[0223] As an alternative embodiment, see Fig.15 , the method further comprises:

[0224] S1510. Extracting features of the interactive operation information between the object to be identified and the third interactive object to obtain a third interactive operation feature corresponding to the object to be identified; the third interactive object is an object that interacts with the object to be identified based on the object association relationship corresponding to the object to be identified;

[0225] Based on the first operation feature and the second operation feature, the live broadcast operation of the object to be identified is predicted, and the live broadcast prediction information obtained includes:

[0226] S1520. Based on the first operation feature, the second operation feature and the third interactive operation feature, predict the live broadcast operation of the object to be identified to obtain live broadcast prediction information;

[0227] As an optional embodiment, the third interactive operation feature includes a first social operation feature obtained based on social operation information actively performed by the object to be identified, and a second social operation feature obtained based on social operation information passively performed by the object to be identified. The first social operation feature includes the frequency and number of comments, connections, private messages, etc. actively performed by the object to be identified, or other social information, and the second social operation feature includes the frequency and number of comments, connections, private messages, etc. actively performed by an associated object of the object to be identified, or other social information.

[0228] The first social operation feature can reflect the social activity of the object to be identified. When the activity level corresponding to the first social operation feature is higher, it means that the social activity level of the object to be identified is higher, and the relationship between the object to be identified and the associated object in the target application is closer, and the object to be identified is more likely to continue the live broadcast after the live broadcast is interrupted. When the activity level corresponding to the first social operation feature is lower, it means that the social activity level of the object to be identified is lower, and the relationship between the object to be identified and the associated object in the target application is more distant, and the object to be identified is less likely to continue the live broadcast after the live broadcast is interrupted.

[0229] The second social operation feature can reflect the feedback information of the third interactive object on the identified object in social interaction. In the case where the activity level corresponding to the second social operation feature is higher, it means that the social interaction of the object to be identified has received more positive feedback, and the relationship between the object to be identified and the associated object in the target application is closer, so the object to be identified is more likely to continue the live broadcast after the live broadcast is interrupted. In the case where the activity level corresponding to the second social operation feature is lower, it means that the social interaction of the object to be identified has received less positive feedback, and the relationship between the object to be identified and the associated object in the target application is more distant, so the object to be identified is less likely to continue the live broadcast after the live broadcast is interrupted.

[0230] When predicting the live broadcast operation, features related to the social relationship of the object to be identified can be added. Through the social relationship of the object to be identified, the degree of dependence of the object to be identified on the target application can be determined, thereby improving the accuracy of the live broadcast operation prediction.

[0231] As an optional embodiment, the first operation feature and the second operation feature can be processed by feature accumulation to obtain the corresponding accumulated operation features of the object to be identified in the historical time. The accumulated operation features can be the total information of various operation features in the historical time when the object to be identified performs live broadcast, and the total information of various operation features in the historical time when the object to be identified performs content publishing. The accumulated operation features can include the accumulated amount of live broadcast operation, content publishing operation, interactive information of live broadcast or published content, registration days of the target application, following users and following users, etc. of the object to be identified in the historical time. The accumulated operation features can reflect the use of the target application by the object to be identified. When the use of the target application by the object to be identified indicates that the object to be identified is experienced and frequently uses the target application, it means that the object to be identified has a higher possibility of continuing the live broadcast after interrupting the live broadcast. When the use of the target application by the object to be identified indicates that the object to be identified lacks use experience and rarely uses the target application, it means that the object to be identified has a lower possibility of continuing the live broadcast after interrupting the live broadcast.

[0232] As an optional embodiment, the above object recognition method can be applied to the scenario of recalling the anchor during the silent period. The silent period refers to the period when the anchor interrupts the live broadcast operation. The live broadcast information is the live broadcast operation information, and the content publishing information is the content publishing operation information.

[0233] The object to be identified is the silent period anchor object, and the interruption duration information corresponding to the object to be identified that interrupts the live broadcast information is obtained, that is, the silent period duration information. The preset object whose interruption duration information is greater than or equal to the object identification threshold and less than or equal to the interruption duration threshold can be used as the object to be identified.

[0234] By extracting features of the operation information related to the live broadcast, the operation information related to the content publishing operation, and the interactive operation information in the first time period, the live broadcast operation features, the first content publishing operation features, and the first interactive operation features can be obtained. The live broadcast operation features, the first content publishing operation features, and the first interactive operation features all include operation features in two dimensions: active and passive. The more active and positive feedback the object to be identified is determined based on the live broadcast operation features, the first content publishing operation features, and the first interactive operation features, the higher the possibility that the object to be identified will continue to broadcast live.

[0235] By extracting features from the operation information and interactive operation information related to the content publishing operation in the second time period, a second content publishing operation feature and a second interactive operation feature can be obtained. The second content publishing operation feature and the second interactive operation feature both include operation features in two dimensions: active and passive. The more active and positive feedback the object to be identified based on the second content publishing operation feature and the second interactive operation feature, the lower the possibility that the object to be identified will continue to live broadcast.

[0236] The first operation feature within the first time period is aggregated and accumulated, and the second operation feature within the second time period is aggregated and accumulated, so that the accumulated operation feature can be obtained.

[0237] The first time period and the second time period are divided by the interruption time node of the live broadcast information interruption execution. According to the timing information corresponding to the first time period, the number of operations of the live broadcast feature in each first sub-time period is determined, and based on the change of the number of operations, the change trend of the live broadcast feature in the first time period can be determined, and the live broadcast operation trend information can be obtained. According to the timing information corresponding to the second time period, the number of operations of the content publishing feature in each second sub-time period is determined, and based on the change of the number of operations, the change trend of the content publishing feature in the second time period is determined, and the content publishing operation trend information can be obtained.

[0238] For the live broadcast operation feature in the first time period, based on the distance between the time interval corresponding to the live broadcast operation feature and the interruption time node, the weight information corresponding to the live broadcast operation feature can be determined. Based on the weight information, the live broadcast operation feature is weighted to obtain the live broadcast operation timing feature, and the live broadcast operation timing feature is accumulated to obtain the target content production feature corresponding to the live broadcast information in the first time period.

[0239] The live broadcast operation time sequence feature, the first operation feature, the second operation feature, the operation trend information, the cumulative operation feature and the social operation feature are input into the preset object recognition model for object recognition, and the target object that continues the live broadcast within the preset time period can be determined from the objects to be recognized. The target object is the anchor object that may continue the live broadcast after the object is recalled within the preset time period. Object recall information can be sent to the target object, the target object is recalled, and the recall processing result is obtained, that is, whether the target object continues the live broadcast.

[0240] The disclosed embodiment proposes an object recognition method, which includes: extracting features from the operation information of the object to be recognized within a first time period and the operation information within a second time period, and obtaining first operation features and second operation features corresponding to the object to be recognized. The object to be recognized is an object that has not been broadcast live for N consecutive days, and the first time period is the time period before the N days; N days include the current time, and N is an integer greater than 1. Based on the first operation feature and the second operation feature, the live broadcast operation of the object to be recognized is predicted to obtain live broadcast prediction information. The method integrates the features corresponding to the live broadcast information of the object to be recognized before the live broadcast is interrupted, and the features corresponding to the activity level of the object to be recognized on the target application after the live broadcast is interrupted, and identifies potential target objects by integrating multi-dimensional features, thereby improving the accuracy and effectiveness of object recognition, and also improving the interactive experience between the object to be recognized and the interactive object, forming positive interactive feedback.

[0241] Fig.16 is a block diagram of an object recognition device according to an exemplary embodiment. Fig.16 , the device comprises:

[0242] The first feature extraction module 1610 is configured to perform feature extraction on the operation information of the object to be identified within a first time period to obtain a first operation feature corresponding to the object to be identified; the object to be identified is an object that has not performed a live broadcast operation for N consecutive days, and the first time period is a time period before N days; N days includes the current time, and N is an integer greater than 1;

[0243] The second feature extraction module 1620 is configured to perform feature extraction on the operation information of the object to be identified in a second time period to obtain a second operation feature corresponding to the object to be identified; the second time period is a time period after the first time period;

[0244] The live broadcast prediction module 1630 is configured to predict the live broadcast operation of the object to be identified based on the first operation feature and the second operation feature to obtain live broadcast prediction information;

[0245] The target object determination module 1640 is configured to determine the target object for continuing to perform the live broadcast operation from the objects to be identified based on the live broadcast prediction information.

[0246] As an optional embodiment, the first operation feature includes a live broadcast operation feature, and the live broadcast prediction module includes:

[0247] A weighted processing unit is configured to perform weighted processing on the live broadcast operation feature based on weight information corresponding to the live broadcast operation feature to obtain a live broadcast operation time sequence feature;

[0248] A live broadcast operation trend determination unit is configured to determine a change trend of the first operation feature within the first time period according to the time sequence information corresponding to the first time period, and obtain live broadcast operation trend information;

[0249] a content publishing operation trend determining unit configured to determine a change trend of the second operation feature within the second time period according to the time sequence information corresponding to the second time period, and obtain content publishing operation trend information;

[0250] The live broadcast operation prediction unit is configured to predict the live broadcast operation of the object to be identified based on the first operation feature, the second operation feature, the live broadcast operation timing feature, the live broadcast operation trend information and the content publishing operation trend information to obtain the live broadcast prediction information.

[0251] As an optional embodiment, the weighted processing unit includes:

[0252] The weight information determining unit is configured to determine the weight information corresponding to the live broadcast operation feature based on the distance between the time node corresponding to the live broadcast operation feature in the first time period and the interruption time node; the interruption time node is the time node dividing the first time period and the second time period; the weight information is information that increases as the distance decreases;

[0253] The timing feature determination unit is configured to perform weighted processing on the live broadcast operation features based on the weight information to obtain the live broadcast operation timing features.

[0254] As an optional embodiment, the first operation feature includes a live broadcast operation feature, a first content publishing operation feature, and a first interactive operation feature, and the first feature extraction module includes:

[0255] A live broadcast operation feature extraction unit is configured to perform feature extraction on operation information related to the live broadcast operation of the object to be identified within the first time period to obtain a live broadcast operation feature;

[0256] A first content publishing operation feature extraction unit is configured to perform feature extraction on operation information related to the content publishing operation of the object to be identified within a first time period to obtain a first content publishing operation feature;

[0257] The first interactive operation feature extraction unit is configured to perform feature extraction on the interactive operation information performed by the to-be-identified object within a first time period to obtain a first interactive operation feature.

[0258] As an optional embodiment, the live broadcast operation feature includes a live broadcast feature corresponding to the object to be identified and a live broadcast association feature corresponding to the first associated object, and the first associated object is an object having an association relationship with the object to be identified within the first time period;

[0259] The live broadcast operation feature extraction unit includes:

[0260] A live broadcast feature extraction unit is configured to perform feature extraction on live broadcast operation information performed by the object to be identified within a first time period to obtain live broadcast features;

[0261] The live broadcast associated feature extraction unit is configured to perform feature extraction on first associated operation information performed by the first associated object within the first time period to obtain live broadcast associated features, where the first associated operation information is operation information associated with the production content corresponding to the live broadcast operation information.

[0262] As an optional embodiment, the first content publishing operation feature includes a first content publishing feature corresponding to the object to be identified and a first content publishing association feature corresponding to the first associated object, and the first associated object is an object having an association relationship with the object to be identified within the first time period;

[0263] The first content publishing operation feature extraction unit includes:

[0264] A first content publishing feature extraction unit is configured to perform feature extraction on content publishing operation information performed by the object to be identified within a first time period to obtain a first content publishing feature;

[0265] The first content release associated feature extraction unit is configured to perform feature extraction on the second associated operation information performed by the first associated object within the first time period to obtain the first content release associated feature; the second associated operation feature is the operation information associated with the production content corresponding to the content release operation information within the first time period.

[0266] As an optional embodiment, the first interactive operation feature includes a first interactive feature corresponding to the object to be identified and a second interactive feature corresponding to the first interactive object, and the first interactive object is an object that interacts with the object to be identified based on the production content of the object to be identified within the first time period;

[0267] The first interactive operation feature extraction unit includes:

[0268] A first interaction feature extraction unit is configured to extract features of interaction operation information performed by the object to be identified within a first time period to obtain a first interaction feature;

[0269] The second interaction feature extraction unit is configured to perform feature extraction on the interaction operation information performed by the first interaction object within the first time period to obtain a second interaction feature.

[0270] As an optional embodiment, the second operation feature includes a second content publishing operation feature and a second interactive operation feature, and the second feature extraction module includes:

[0271] A second content publishing operation feature extraction unit is configured to perform feature extraction on content publishing operation information performed by the object to be identified within a second time period to obtain a second content publishing operation feature;

[0272] The second interactive operation feature extraction unit is configured to perform feature extraction on the interactive operation information performed by the to-be-identified object within a second time period to obtain a second interactive operation feature.

[0273] As an optional embodiment, the second content publishing operation feature includes a second content publishing feature corresponding to the object to be identified and a content publishing association feature corresponding to a second associated object, where the second associated object is an object having an association relationship with the object to be identified within the second time period;

[0274] The second content publishing operation feature extraction unit includes:

[0275] A second content publishing feature extraction unit is configured to extract features of content publishing operation information performed by the object to be identified within a second time period to obtain a second content publishing feature;

[0276] The second content release associated feature extraction unit is configured to perform feature extraction on the third associated operation information performed by the second associated object within the second time period to obtain the second content release associated feature; the third associated operation feature is the operation information associated with the production content corresponding to the content release operation information within the second time period.

[0277] As an optional embodiment, the second interactive operation feature includes a third interactive feature corresponding to the object to be identified and a fourth interactive feature corresponding to the second interactive object, and the second interactive object is an object that interacts with the object to be identified based on the production content of the object to be identified within the second time period;

[0278] The second interactive operation feature extraction unit includes:

[0279] A third interaction feature extraction unit is configured to perform feature extraction on the interaction operation information performed by the object to be identified within the second time period to obtain a third interaction feature;

[0280] The fourth interaction feature extraction unit is configured to perform feature extraction on the interaction operation information performed by the second interaction object within the second time period to obtain a fourth interaction feature.

[0281] As an optional embodiment, the device further includes:

[0282] A third interactive operation feature extraction module is configured to perform feature extraction on interactive operation information between the object to be identified and a third interactive object to obtain a third interactive operation feature corresponding to the object to be identified; the third interactive object is an object that interacts with the object to be identified based on an object association relationship corresponding to the object to be identified;

[0283] The live broadcast prediction module includes:

[0284] The live broadcast prediction unit is configured to predict the live broadcast operation of the object to be identified based on the first operation feature, the second operation feature and the third interactive operation feature to obtain live broadcast prediction information.

[0285] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0286] Fig.17 is a block diagram of an electronic device for object recognition according to an exemplary embodiment. The electronic device may be a server, and its internal structure diagram may be as shown in FIG. Fig.17 As shown. The electronic device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the electronic device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, an object recognition method is implemented.

[0287] Those skilled in the art will understand that Fig.17 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present disclosure, and does not constitute a limitation on the electronic device to which the scheme of the present disclosure is applied. The specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0288] In an exemplary embodiment, an electronic device is also provided, including: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the object recognition method in the embodiment of the present disclosure.

[0289] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory including instructions, and the above instructions can be executed by a processor of an electronic device to perform the above method. Optionally, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0290] In an exemplary embodiment, a computer program product including instructions is also provided. When the computer program product is run on a computer, the computer is enabled to execute the image defect filling network determination method or the image defect processing method in the embodiments of the present disclosure.

[0291] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0292] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0293] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for object recognition, It is characterized in that The method comprises: Extracting features from the operation information of the object to be identified within a first time period to obtain a first operation feature corresponding to the object to be identified; the object to be identified is an object that has not performed a live broadcast operation for N consecutive days, and the first time period is a time period before the N days; the N days include the current time, and N is an integer greater than 1; Extracting features of the operation information of the object to be identified within a second time period to obtain a second operation feature corresponding to the object to be identified; the second time period is a time period after the first time period; Based on the first operation feature and the second operation feature, predicting the live broadcast operation of the to-be-identified object to obtain live broadcast prediction information; Based on the live broadcast prediction information, determining a target object for continuing to perform a live broadcast operation from the to-be-identified objects; the first operation feature includes a live broadcast operation feature; The predicting the live broadcast operation of the to-be-identified object based on the first operation feature and the second operation feature to obtain the live broadcast prediction information includes: Based on the weight information corresponding to the live broadcast operation feature, weighted processing is performed on the live broadcast operation feature to obtain a live broadcast operation time sequence feature; Determine, according to the time sequence information corresponding to the first time period, a change trend of the first operation feature within the first time period to obtain live broadcast operation trend information; Determining, according to the time series information corresponding to the second time period, a change trend of the second operation feature within the second time period to obtain content publishing operation trend information; Based on the first operation feature, the second operation feature, the live broadcast operation timing feature, the live broadcast operation trend information and the content publishing operation trend information, the live broadcast operation of the object to be identified is predicted to obtain live broadcast prediction information.

2. The object recognition method according to claim 1, It is characterized in that The weighting process of the live broadcast operation feature based on the weight information corresponding to the live broadcast operation feature to obtain the live broadcast operation timing feature includes: Determine the weight information corresponding to the live broadcast operation feature based on the distance between the time node corresponding to the live broadcast operation feature in the first time period and the interruption time node; the interruption time node is the time node dividing the first time period and the second time period; the weight information is information that increases as the distance decreases; Based on the weight information, the live broadcast operation characteristics are weighted to obtain the live broadcast operation timing characteristics.

3. The object recognition method according to any one of claims 1 to 2, It is characterized in that The first operation feature includes a live broadcast operation feature, a first content publishing operation feature, and a first interactive operation feature. The feature extraction of the operation information of the object to be identified within the first time period to obtain the first operation feature corresponding to the object to be identified includes: Extracting features of operation information related to the live broadcast operation of the object to be identified within the first time period to obtain the live broadcast operation features; Extracting features of operation information related to content publishing operations of the to-be-identified object within the first time period to obtain the first content publishing operation features; Feature extraction is performed on the interactive operation information performed by the object to be identified within the first time period to obtain the first interactive operation feature.

4. The object recognition method according to claim 3, It is characterized in that The live broadcast operation feature includes a live broadcast feature corresponding to the object to be identified and a live broadcast association feature corresponding to a first associated object, wherein the first associated object is an object having an association relationship with the object to be identified within the first time period; The extracting of features of the operation information related to the live broadcast operation of the object to be identified within the first time period to obtain the live broadcast operation features includes: Extracting features of live broadcast operation information performed by the to-be-identified object within the first time period to obtain the live broadcast features; Feature extraction is performed on first associated operation information performed by the first associated object within the first time period to obtain the live broadcast associated feature, where the first associated operation information is operation information associated with the production content corresponding to the live broadcast operation information.

5. The object recognition method according to claim 3, It is characterized in that The first content publishing operation feature includes a first content publishing feature corresponding to the object to be identified and a first content publishing association feature corresponding to a first associated object, wherein the first associated object is an object having an association relationship with the object to be identified within the first time period; The extracting features of the operation information related to the content publishing operation of the to-be-identified object within the first time period to obtain the first content publishing operation features includes: Extracting features of content publishing operations performed by the to-be-identified object within the first time period to obtain the first content publishing features; Feature extraction is performed on second associated operation information performed by the first associated object within the first time period to obtain the first content publishing associated feature; the second associated operation information is operation information associated with the production content corresponding to the content publishing operation information within the first time period.

6. The object recognition method according to claim 3, It is characterized in that The first interactive operation feature includes a first interactive feature corresponding to the object to be identified and a second interactive feature corresponding to a first interactive object, the first interactive object being an object that interacts with the object to be identified based on the content produced by the object to be identified within the first time period; The extracting features of the interactive operation information performed by the to-be-identified object within the first time period to obtain the first interactive operation features includes: Extracting features of interactive operation information performed by the object to be identified within the first time period to obtain the first interactive features; Feature extraction is performed on the interactive operation information performed by the first interactive object in the first time period to obtain the second interactive feature.

7. The object recognition method according to any one of claims 1 to 2, It is characterized in that The second operation feature includes a second content publishing operation feature and a second interactive operation feature, and the feature extraction of the operation information of the object to be identified within the second time period to obtain the second operation feature corresponding to the object to be identified includes: Extracting features of the content publishing operation information performed by the to-be-identified object within the second time period to obtain the second content publishing operation features; Feature extraction is performed on the interactive operation information performed by the object to be identified within the second time period to obtain the second interactive operation feature.

8. The object recognition method according to claim 7, It is characterized in that The second content publishing operation feature includes a second content publishing feature corresponding to the object to be identified and a content publishing association feature corresponding to a second associated object, wherein the second associated object is an object having an association relationship with the object to be identified within the second time period; The extracting features of the operation information related to the content publishing operation of the to-be-identified object within the second time period to obtain the second content publishing operation features includes: Extracting features of the content publishing operation information performed by the to-be-identified object within the second time period to obtain the second content publishing features; Feature extraction is performed on the third associated operation information performed by the second associated object within the second time period to obtain content publishing associated features corresponding to the second associated object; the third associated operation information is operation information associated with the production content corresponding to the content publishing operation information within the second time period.

9. The object recognition method according to claim 7, It is characterized in that The second interactive operation feature includes a third interactive feature corresponding to the object to be identified and a fourth interactive feature corresponding to a second interactive object, and the second interactive object is an object that interacts with the object to be identified based on the content produced by the object to be identified within the second time period; The extracting features of the interactive operation information performed by the to-be-identified object within the second time period to obtain the second interactive operation features includes: Extracting features of interactive operation information performed by the object to be identified within the second time period to obtain the third interactive feature; Feature extraction is performed on the interactive operation information performed by the second interactive object in the second time period to obtain the fourth interactive feature.

10. The object recognition method according to claim 1, It is characterized in that The method further comprises: Extracting features of interactive operation information between the object to be identified and a third interactive object to obtain a third interactive operation feature corresponding to the object to be identified; the third interactive object is an object that interacts with the object to be identified based on an object association relationship corresponding to the object to be identified; The predicting the live broadcast operation of the to-be-identified object based on the first operation feature and the second operation feature to obtain the live broadcast prediction information includes: Based on the first operation feature, the second operation feature and the third interactive operation feature, the live broadcast operation of the object to be identified is predicted to obtain live broadcast prediction information.

11. An object recognition device, It is characterized in that The device comprises: A first feature extraction module is configured to perform feature extraction on operation information of an object to be identified within a first time period to obtain a first operation feature corresponding to the object to be identified; the object to be identified is an object that has not performed a live broadcast operation for N consecutive days, and the first time period is a time period before the N days; the N days include the current time, and N is an integer greater than 1; A second feature extraction module is configured to perform feature extraction on the operation information of the object to be identified within a second time period to obtain a second operation feature corresponding to the object to be identified; the second time period is a time period after the first time period; A live broadcast prediction module is configured to predict the live broadcast operation of the to-be-identified object based on the first operation feature and the second operation feature to obtain live broadcast prediction information; A target object determination module is configured to determine a target object for continuing to perform a live broadcast operation from the to-be-identified object based on the live broadcast prediction information; the first operation feature includes a live broadcast operation feature; The predicting the live broadcast operation of the to-be-identified object based on the first operation feature and the second operation feature to obtain the live broadcast prediction information includes: Based on the weight information corresponding to the live broadcast operation feature, weighted processing is performed on the live broadcast operation feature to obtain a live broadcast operation time sequence feature; Determine, according to the time sequence information corresponding to the first time period, a change trend of the first operation feature within the first time period to obtain live broadcast operation trend information; Determining, according to the time series information corresponding to the second time period, a change trend of the second operation feature within the second time period to obtain content publishing operation trend information; Based on the first operation feature, the second operation feature, the live broadcast operation timing feature, the live broadcast operation trend information and the content publishing operation trend information, the live broadcast operation of the object to be identified is predicted to obtain live broadcast prediction information.

12. An electronic device, It is characterized in that The electronic device comprises: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the object recognition method according to any one of claims 1 to 10.

13. A computer-readable storage medium, It is characterized in that When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the object recognition method according to any one of claims 1 to 10.