Content request method and device, electronic device and storage medium

Abstract

The present disclosure relates to a content request method and apparatus, an electronic device, and a storage medium. The content request method comprises: determining current state information of a terminal device; inputting the current state information into a neural network model running within the terminal device to obtain recommended parameters for requesting a target content from a content server; and sending a content request containing the recommended parameters to the content server.

Description

Technical Field

[0001] The present disclosure relates to the field of information technology, and in particular to a content request method and device, an electronic device, and a storage medium. Background Art

[0002] Mobile phones, bracelets, smart watches, or car-mounted devices are daily terminal devices that are used frequently by users and are often used by users for work, study, or entertainment.

[0003] When used for work, study, or entertainment, these terminal devices interact with the network. Typically, the terminal device sends a request to the network, which then returns the requested content based on the request. However, the parameters included in the request sent by the terminal device may be default or random. Alternatively, the network server returns the requested content to the terminal device based on parameters determined by itself. Summary of the Invention

[0004] Embodiments of the present disclosure provide a content request method and device, an electronic device, and a storage medium.

[0005] A first aspect of an embodiment of the present disclosure provides a content request method, including:

[0006] Determine current status information of the terminal device;

[0007] Inputting the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting target content from a content server;

[0008] Sending a content request carrying the suggested parameters to the content server.

[0009] Based on the above solution, the step of inputting the current state information into the neural network model running in the terminal device includes:

[0010] Inputting the current state information and the nth set of candidate parameters into the neural network model to obtain predicted state information; wherein the predicted state information indicates: a state that the terminal device is ready to enter when requesting target content with the nth candidate parameter; n is a positive integer less than N; N is the total number of sets of candidate parameters; and a set of candidate parameters includes at least one request parameter;

[0011] When the state that the terminal device is ready to enter meets a preset condition, the nth set of candidate parameters is used as the recommended parameters.

[0012] Based on the above solution, the method further includes:

[0013] Calculating a quality score value based on the state of the terminal device to be entered and the nth set of candidate parameters;

[0014] The terminal ready-to-enter state with the largest quality score value is determined as the terminal ready-to-enter state that meets the preset condition.

[0015] Based on the above solution, the current status information includes at least one of the following:

[0016] Network status information of the terminal device;

[0017] Remaining power information of the terminal device;

[0018] location information of the terminal device;

[0019] time information when the terminal device requests the target content;

[0020] The terminal device currently outputs content parameters of the target content.

[0021] Based on the above solution, the network status information indicates at least one of the following:

[0022] The network provider to which the terminal device is currently connected;

[0023] The type of network to which the terminal device is currently connected;

[0024] The current network signal strength of the terminal device;

[0025] The current network bandwidth of the terminal device.

[0026] Based on the above solution, the method further includes:

[0027] Determining whether the terminal device stores model information of the neural network model that has completed at least one round of local training;

[0028] The step of inputting the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting a target content from a content server includes:

[0029] When the terminal device stores model information of the neural network model that has completed at least one round of local training, running the neural network model according to the model information;

[0030] The current state information is input into the neural network model running locally on the terminal device to obtain the recommended parameters.

[0031] Based on the above solution, the method further includes:

[0032] When the terminal device does not store the model information of the neural network model that has completed at least one round of local training, the recommended parameters are determined according to the default strategy and the current state information.

[0033] Based on the above solution, determining the recommended parameters according to the default policy and the current state information includes:

[0034] When the network type connected to the terminal device is a non-traffic sensitive network, the remaining power of the terminal device is greater than a power threshold, and / or the network signal strength connected to the terminal device is greater than a strength threshold, determining the mth set of alternative parameters as the recommended parameters; otherwise, determining the m+1th set of alternative parameters as the recommended parameters;

[0035] The content quality of the target content output by the terminal device using the mth set of candidate parameters is higher than the content quality of the target content output by the terminal device using the (m+1)th set of candidate parameters.

[0036] Based on the above scheme, the alternative parameters also include: the m+2th set of alternative parameters, wherein the content quality when the terminal device outputs the target content with the m+2th set of alternative parameters is higher than the content quality when the terminal device outputs the target content with the m+1th set of alternative parameters.

[0037] Based on the above solution, the method further includes:

[0038] storing the state information of the terminal device before requesting the target content, the request parameters of the target content, and the state information of the terminal device after requesting the target content in a history database;

[0039] When the number of information items in the historical database reaches a preset number, the neural network model is subjected to a round of local training using the historical data recorded in the historical database as sample data.

[0040] Based on the above solution, the historical database includes a first-in-first-out queue, wherein the status information of the terminal device before the target content request, the request parameters of the target content, and the status information after the terminal device requests the target content are correspondingly stored in the historical database, including:

[0041] When the number of information items in the history database reaches a preset number, the status information of the terminal device before the target content request, the request parameters of the target content, and the status information after the terminal device requests the target content are stored at the end of the first-in-first-out queue.

[0042] A second aspect of an embodiment of the present disclosure provides a content request device, including:

[0043] A first determining module, configured to determine current status information of a terminal device;

[0044] a parameter module, configured to input the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting a target content from a content server;

[0045] The request module is configured to send a content request carrying the suggested parameters to the content server.

[0046] Based on the above scheme, the parameter module is specifically used to input the current state information and the nth set of alternative parameters into the neural network model to obtain predicted state information; wherein, the predicted state information indicates: the state that the terminal device is ready to enter when requesting the target content with the nth alternative parameter; n is a positive integer less than N; N is the total number of sets of alternative parameters; a set of the alternative parameters includes at least one request parameter; when the state that the terminal device is ready to enter meets the preset conditions, the nth set of alternative parameters is used as the recommended parameters.

[0047] Based on the above solution, the device further includes:

[0048] a calculation module, configured to calculate a quality score value based on the state of the terminal device to be entered and the nth set of candidate parameters;

[0049] The determination module is configured to determine the terminal ready-to-enter state with the largest quality score as the terminal ready-to-enter state that meets the preset condition.

[0050] Based on the above solution, the current status information includes at least one of the following:

[0051] Network status information of the terminal device;

[0052] Remaining power information of the terminal device;

[0053] location information of the terminal device;

[0054] time information when the terminal device requests the target content;

[0055] The terminal device currently outputs content parameters of the target content.

[0056] Based on the above solution, the network status information indicates at least one of the following:

[0057] The network provider to which the terminal device is currently connected;

[0058] The type of network to which the terminal device is currently connected;

[0059] The current network signal strength of the terminal device;

[0060] The current network bandwidth of the terminal device.

[0061] Based on the above solution, the device further includes:

[0062] A second determination module is configured to determine whether the terminal device stores model information of the neural network model that has completed at least one round of local training;

[0063] The parameter module is specifically used to run the neural network model according to the model information when the terminal device stores the model information of the neural network model that has completed at least one round of local training; input the current state information into the neural network model running locally on the terminal device to obtain the recommended parameters.

[0064] Based on the above solution, the parameter module is further specifically used to determine the recommended parameters according to the default strategy and the current status information when the terminal device does not store the model information of the neural network model that has completed at least one round of local training.

[0065] Based on the above scheme, the parameter module is specifically used to determine the mth set of alternative parameters as the recommended parameters when the network type connected to the terminal device is a non-traffic sensitive network, the remaining power of the terminal device is greater than the power threshold and / or the network signal strength connected to the terminal device is greater than the strength threshold; otherwise, determine the m+1th set of alternative parameters as the recommended parameters;

[0066] The content quality of the target content output by the terminal device using the mth set of candidate parameters is higher than the content quality of the target content output by the terminal device using the (m+1)th set of candidate parameters.

[0067] Based on the above scheme, the alternative parameters also include: the m+2th set of alternative parameters, wherein the content quality when the terminal device outputs the target content with the m+2th set of alternative parameters is higher than the content quality when the terminal device outputs the target content with the m+1th set of alternative parameters.

[0068] Based on the above solution, the device further includes:

[0069] a storage module, configured to store the state information of the terminal device before the target content is requested, the request parameters of the target content, and the state information after the terminal device requests the target content, in a history database;

[0070] The training module is used to perform a round of local training on the neural network model using the historical data recorded in the historical database as sample data when the number of information items in the historical database reaches a preset number.

[0071] Based on the above scheme, the historical database includes: a first-in-first-out queue, wherein the storage module is specifically used to store the status information of the terminal device before the target content request, the request parameters of the target content, and the status information after the terminal device requests the target content to the end of the first-in-first-out queue when the number of information items in the historical database reaches a preset number.

[0072] A third aspect of the present disclosure provides an electronic device, including:

[0073] a memory for storing processor-executable instructions;

[0074] a processor connected to the memory;

[0075] The processor is configured to execute the content request method provided by any technical solution of the first aspect.

[0076] A fourth aspect of the embodiments of the present disclosure provides a non-transitory computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are executed by a processor, the content request method provided by any technical solution of the first aspect is implemented.

[0077] The technical solutions provided by the embodiments of the present disclosure may have the following beneficial effects:

[0078] The terminal device will run a neural network model locally. Before the terminal device sends a content request, it will calculate recommended parameters based on the neural network model, based on the terminal device's current state information. These parameters will be sent to the server in the content request, and the target content will be requested. This will ensure that the target content is suitable for the terminal device's current state, balancing the output quality of the target content with the overall user experience of the terminal device. Furthermore, the neural network model runs locally on the terminal device, providing a personalized model for the terminal device rather than a general model running on the network. This allows it to better provide recommended parameters suitable for the current state based on the characteristics of different terminal devices and their state information at a specific time, resulting in high accuracy and personalization.

[0079] 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 disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0081] Figure 1 It is a flowchart of a content request method shown in an embodiment of the present disclosure.

[0082] Figure 2 It is a flowchart of a content request method shown in an embodiment of the present disclosure.

[0083] Figure 3 It is a flowchart of a content request method shown in an embodiment of the present disclosure.

[0084] Figure 4 It is a flowchart of a content request method shown in an embodiment of the present disclosure.

[0085] Figure 5 It is a flowchart of a content request method shown in an embodiment of the present disclosure.

[0086] Figure 6 It is a flowchart of a content request method shown in an embodiment of the present disclosure.

[0087] Figure 7 It is a schematic diagram of the training process of a neural network model shown in an embodiment of the present disclosure.

[0088] Figure 8 It is a flowchart of a content request method shown in an embodiment of the present disclosure.

[0089] Figure 9 It is a structural diagram of a content request device shown in an embodiment of the present disclosure.

[0090] Figure 10 It is a structural diagram of an electronic device shown in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0091] Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. In the following description, when referring to the drawings, like numbers in different figures represent like or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present invention. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present invention, as detailed in the appended claims.

[0092] like Figure 1 As shown, the embodiment of the present disclosure provides a content request method, including:

[0093] S110: Determine current status information of the terminal device;

[0094] S120: Inputting the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting a target content from a content server;

[0095] S130: Sending a content request carrying the suggested parameters to the content server.

[0096] The content request method provided by the embodiments of the present disclosure can be applied to a terminal device, which includes but is not limited to an electronic device running various clients.

[0097] The electronic device includes but is not limited to: a mobile phone, a tablet computer, a wearable device or a vehicle-mounted device.

[0098] The current status information here may include at least one of the following:

[0099] Current network status information of the terminal device;

[0100] The current operating status information of the terminal device.

[0101] The current network status information indicates the current status of the network to which the terminal device is connected;

[0102] The current operating status information indicates the current operating status of the terminal device.

[0103] The current status of the network to which the terminal device is connected includes but is not limited to at least one of the following:

[0104] The network type to which the terminal device is connected, for example, whether it is connected to a cellular mobile communication network, a WiFi network, or a wired network. If it is connected to a cellular mobile communication network, is it a 5G network or a 4G network?

[0105] the network signal strength of the network to which the terminal device is connected;

[0106] The number of networks to which the terminal device is connected. Sometimes, a terminal device is connected to a cellular mobile communication and / or WiFi network.

[0107] The current operating status includes but is not limited to at least one of the following:

[0108] The current load rate of the terminal equipment;

[0109] The current remaining available buffer capacity of the terminal device;

[0110] The current number of idle threads of the terminal device.

[0111] In an embodiment of the present disclosure, the terminal device locally runs a neural network module, and this neural network model is a deep learning model trained with sample data.

[0112] The neural network model running locally on the terminal device takes the current state information of the terminal device as input. After a series of operations in the hidden layer of the neural network model, there will be an output result in the output layer. The output result can directly be the recommended parameters or the basis for determining the recommended parameters.

[0113] The suggested parameters may be various parameters that the terminal device uses to request the target content from the server.

[0114] In some embodiments, the suggested parameters include but are not limited to various parameters that can determine the data volume of the target content.

[0115] Exemplarily, the suggested parameters include but are not limited to at least one of the following:

[0116] If the target content is a video, the suggested parameters may include: video clarity or video image resolution;

[0117] If the target content is an image, the recommended parameter may be: image resolution; if the target content can also be any multimedia information, the recommended parameter may also be version; different versions have the same content but different data volumes. For example, if the target content is audio, audios of different sound qualities may have the same audio content but different sound quality.

[0118] Suggested parameters are given by the neural network model running locally on the terminal device, without the need for a server to give suggested parameters, thus sharing the server's load rate; at the same time, the neural network model running on the terminal device can be trained based on the terminal device's own historical status information and / or the historical request status of the content, so that it can adapt to the terminal device itself.

[0119] Different terminal devices have different software and hardware conditions and / or network conditions. If the neural network model is run locally in the terminal device, the predicted recommended parameters will be personalized parameters adapted to the terminal device, thereby improving the service quality of the target content while improving the service quality of the terminal device and improving user satisfaction.

[0120] In some embodiments, inputting the current state information into a neural network model running in the terminal device includes:

[0121] Inputting the current state information and the nth set of candidate parameters into the neural network model to obtain predicted state information; wherein the predicted state information indicates: a state that the terminal device is ready to enter when requesting target content with the nth candidate parameter; n is a positive integer less than N; N is the total number of sets of candidate parameters; and a set of candidate parameters includes at least one request parameter;

[0122] When the state that the terminal device is ready to enter meets a preset condition, the nth set of candidate parameters is used as the recommended parameters.

[0123] The target content is a video for illustration purposes; the video may include two or three sets of candidate parameters.

[0124] For example, if there are three sets of video parameters, the video definition includes high definition, standard definition, and low definition. For the same video, the data volume of low definition video is smaller than that of standard definition video, and the data volume of standard definition video is smaller than that of high definition video.

[0125] For another example, if there are four sets of optional parameters for a video, the video definition may include: 4K, HD, SD, and LD. For the same video, the data volume of a 4K video is larger than that of an HD video.

[0126] If the current network status of the terminal device is not good and it still requests HD or 4K video, the congestion of the terminal device may be aggravated, and the single frame data of the video will also be very large, resulting in lag in the video playback of the terminal device.

[0127] Based on this, the neural network model running locally on the terminal device will predict recommended parameters that will prevent the terminal device from experiencing lag or increased congestion.

[0128] The suggested parameters are carried in the content request and sent to the server. In this way, the server will return the target content to the terminal device according to the suggested parameters carried in the content request, so as to improve the service quality of the terminal.

[0129] In the disclosed embodiment, the neural network model uses current state information and various sets of candidate parameters as input to predict the state change of the terminal device when requesting target content using various sets of candidate parameters under the terminal device state indicated by the current state information, thereby predicting the state the terminal device will enter. Furthermore, the model determines whether the state the terminal device will briefly enter under each set of candidate parameters meets preset conditions. If so, the corresponding set of candidate parameters can be used as suggested parameters and included in the content request and sent to the server.

[0130] Exemplarily, the state that the terminal device is ready to enter satisfies a preset condition, including:

[0131] If one or more metrics of the current operating state of the terminal device are within a preset range, and one or more metrics of the state that the device is about to enter predicted by the neural network model do not deteriorate, then it is considered that the state that the device is about to enter meets the preset conditions;

[0132] If one or more metrics of the terminal device's current operating state are lower than a preset range, and one or more metrics of the state the device is about to enter predicted by the neural network model fall within a preset range, then the state the device is about to enter can be considered to meet the preset conditions;

[0133] If one or more measurement indicators of the current operating state of the terminal device are lower than the preset range, and one or more indicators of the state that the device is about to enter predicted by the neural network model are optimized to be better than the indicators corresponding to the preset state, then it can be considered that the state entered by the device meets the preset conditions.

[0134] Of course, the above is merely an example of satisfying the preset conditions for the state that the device is about to enter, and the specific implementation is not limited to this example. For example, the state that the device is about to enter satisfying the preset conditions may include: one or more indicators of the device's upcoming state reaching the indicator values ​​specified by the preset conditions. Specifically, in the state that the terminal device is about to enter, if the frame rate of the received video reaches the frame rate required for broadcast, the preset conditions can be considered to have been met.

[0135] In some embodiments, the method further comprises:

[0136] Calculating a quality score value based on the state of the terminal device to be entered and the nth set of candidate parameters;

[0137] The terminal ready-to-enter state with the largest quality score value is determined as the terminal ready-to-enter state that meets the preset condition.

[0138] In the embodiment of the present disclosure, a quality score value is introduced to quantitatively evaluate the device preparation state after the terminal device requests the target content with the nth set of candidate parameters.

[0139] For example, the quality score is positively correlated with the local service quality of the target content on the terminal device, and negatively correlated with the deterioration of the terminal device's current operating state. Thus, based on the quality score, a balance is selected between the output quality of the target content and the state that the terminal device will enter. This ensures both the service quality of the target content and the user experience of the terminal device to a certain extent.

[0140] In some embodiments, the current status information includes at least one of the following:

[0141] Network status information of the terminal device;

[0142] Remaining power information of the terminal device;

[0143] location information of the terminal device;

[0144] time information when the terminal device requests the target content;

[0145] The terminal device currently outputs content parameters of the target content.

[0146] The network status information of the network to which the terminal device is connected will affect issues such as the bandwidth of the network to which the terminal device is connected and the bandwidth charges.

[0147] The standby time of the terminal device is also very important. Taking video as an example, if the clarity is too high, the power consumption of the video refresh will be too high, thereby shortening the standby time of the terminal device. If the remaining power of the terminal device is already low at this time, this high-definition video is not suitable. At this time, the appropriate recommended parameters can be determined based on the remaining power information of the terminal device.

[0148] The remaining power information may include:

[0149] If it is detected that the terminal device is not in a charging state, the remaining power information indicates the actual remaining power of the terminal device;

[0150] or,

[0151] If it is detected that the terminal device is in a charging state, the remaining power information indicates a preset power value.

[0152] The preset value may be 100%, 95%, or 90%.

[0153] Furthermore, if it is detected that the terminal device is in a wired charging state, the remaining power information indicates a first power value; if it is detected that the terminal device is in a wireless charging state, the remaining power information indicates a second power value; the second power value is less than the first power value.

[0154] Terminal devices at different locations are connected to different types of networks. For example, cellular mobile communication is a traffic-sensitive network. Traffic generated through cellular mobile communication is usually charged, and the unit price of the charge is higher than the unit price of paid WiFi. In this case, even if the signal strength of cellular mobile communication is very strong, you can give priority to whether there is a free WiFi network or a low-cost WiFi network.

[0155] According to the location information of the terminal device, it can be determined whether the terminal device is located in a location such as a company or home where free network access is available.

[0156] The amount of traffic that the network needs to transmit varies in different time periods.

[0157] Moreover, users want different output quality of target content in different time periods.

[0158] Therefore, in the embodiment of the present disclosure, time information will also be considered as the current state information of the terminal device to determine the recommended parameters.

[0159] In some cases, a terminal device continuously requests target content. For example, if a terminal device is currently broadcasting a high-definition video and the video suddenly switches to a low-definition video, the user will notice a noticeable visual difference. Even if the terminal device's video playback remains smooth and data packet transmission and reception are not congested after the switch from high definition to low definition, the user's viewing experience will be poor. Therefore, in these cases, the content parameters of the target content currently being output by the terminal device must be considered to comprehensively determine optimal recommended parameters.

[0160] The current network bandwidth of the terminal device may be the average bandwidth calculated over the current period, or the average bandwidth or minimum bandwidth at the current moment and a period after the current moment predicted based on the network resources allocated by the network side.

[0161] Different bandwidths can transmit different amounts of data per unit time. The larger the bandwidth, the greater the amount of data that can be transmitted per unit time.

[0162] In some embodiments, as Figure 2 As shown, the method further includes:

[0163] S100: Determine whether the terminal device stores model information of the neural network model that has completed at least one round of local training;

[0164] The S120 may include:

[0165] When the terminal device stores model information of the neural network model that has completed at least one round of local training, running the neural network model according to the model information;

[0166] The current state information is input into the neural network model running locally on the terminal device to obtain the recommended parameters.

[0167] If the terminal device has just been put into use or a new application (client) has been downloaded, sufficient sample data for training the neural network model has not yet been collected in the terminal device.

[0168] The initial model parameters of the neural network model may be issued by the network side.

[0169] In one embodiment, the initial model parameters may be trained based on a large amount of data provided by a large number of user devices. In this case, the initial model parameters issued by the network side may be the initial universal model parameters of the neural network. In this case, the neural network model with only the initial model parameters can be directly used to provide recommended parameters.

[0170] In another embodiment, the initial model parameters may be randomly determined parameter values. In this case, the neural network model cannot be directly used for the model with recommended parameters and at least one round of training must be completed before it can be used.

[0171] If the terminal device has collected enough data for one round of training of the neural network model, model training of the neural network model with only initial model parameters begins. For example, after the neural network model completes at least one round of training, the model parameters (e.g., weights or bias values) of the neural network model will change, and such changes enable the neural network to provide appropriate suggested parameters or provide basis parameters for appropriate suggested parameters based on the current state information input.

[0172] In the embodiment of the present disclosure, considering the privacy of the collected terminal device and the phenomenon of heavy network load during training on the network side, the neural network model can be a model trained locally on the terminal device.

[0173] The training of the neural network model can be performed in a preset time period. The preset time period includes but is not limited to at least one of the following:

[0174] Any time period during which the terminal device does not request the target content;

[0175] the midnight period of the time zone where the terminal device is located, etc.;

[0176] The load rate of the terminal device is lower than a preset value during the period.

[0177] In short, the time period for local training of the neural network model is a time period that does not interfere with the terminal device performing other tasks.

[0178] The training of the neural network model is performed in the background of the terminal device and is invisible to the user.

[0179] In other embodiments, the training of the neural network model may be performed by the network. For example, the terminal device submits the current model parameters and training data of the neural network model to the network. After the network completes the training, the updated model parameters are sent to the terminal device. In this way, the terminal device can locally run the neural network model based on the updated model parameters provided by the network.

[0180] That is, after the model parameters of the neural network model are updated, the terminal device will store the model parameters locally, and the terminal device will follow the model parameters when running the neural network again.

[0181] In some embodiments, as Figure 3 As shown, the method further includes:

[0182] S121: When the terminal device does not store the model information of the neural network model that has completed at least one round of local training, determine the recommended parameters according to the default strategy and the current state information.

[0183] The default policy may be automatically configured when the client is installed or updated, may be configured based on user input operations, or may be determined based on historical operations of the terminal device requesting target content.

[0184] In summary, in the embodiment of the present disclosure, if the neural network model stored in the terminal device has not undergone a round of local training, the recommended parameters are determined according to the default strategy.

[0185] The recommended parameters given by the default strategy can also ensure the output quality of the target content and the user experience of the terminal device compared to the randomly given recommended parameters.

[0186] Specifically, determining the recommended parameters according to the default policy and the current state information includes:

[0187] When the network type connected to the terminal device is a non-traffic sensitive network, the remaining power of the terminal device is greater than a power threshold, and / or the network signal strength connected to the terminal device is greater than a strength threshold, determining the mth set of alternative parameters as the recommended parameters; otherwise, determining the m+1th set of alternative parameters as the recommended parameters;

[0188] The content quality of the target content output by the terminal device using the mth set of candidate parameters is higher than the content quality of the target content output by the terminal device using the (m+1)th set of candidate parameters.

[0189] When determining the recommended parameters based on the default strategy, the network type, remaining power, and network signal strength are also taken into consideration to determine the recommended parameters that are suitable for the current situation.

[0190] For example, taking video and image downloading as an example, the mth set of network parameters may be for high-definition video or high-definition image downloading, and the m+1th set of network parameters may be for standard-definition or low-definition video or image downloading.

[0191] In some embodiments, the alternative parameters also include: the m+2th set of alternative parameters, wherein the content quality when the terminal device outputs the target content with the m+2th set of alternative parameters is higher than the content quality when the terminal device outputs the target content with the m+1th set of alternative parameters.

[0192] The content quality may include clarity or image resolution, etc.

[0193] In some other embodiments, the terminal device may have multiple operating modes. For example, the operating modes of the terminal device include: a first mode and a second mode.

[0194] In the first mode, the terminal device locally runs the neural network model and recommends parameters based on the neural network model. In the second mode, the terminal device does not locally run the neural network model. For example, in the second mode, the terminal device may recommend parameters based on a default policy or determine recommended parameters based on historical parameters.

[0195] In some embodiments, the method may include:

[0196] Determine whether the terminal device is located in a preset area. If the terminal device is located in the preset area, determine that the terminal device operates in a first mode. The preset area may be an area where the network status is not very stable.

[0197] Acquire configuration information of the terminal device, and determine an operating mode of the terminal device according to the configuration information. For example, a setting page of the terminal device may receive user input, and generate the configuration information according to the user input;

[0198] A working mode selection prompt is popped up, user input acting on the selection prompt is detected, and the working mode of the terminal device is determined according to the user input acting on the selection prompt.

[0199] In some embodiments, the historical database may be a database that stores a preset number of historical records. After the historical records in the historical database reach a preset number, the terminal device trains the neural network once. After completing a round of training, the historical database can be cleared once.

[0200] In some other embodiments, the historical records stored in the history database may be stored in a first-in-first-out manner, with later historical records overwriting earlier historical records. The terminal device may perform a round of training on the neural network every preset time period.

[0201] In some other embodiments, the method further comprises:

[0202] Counting the accuracy of the state that the device is about to enter as predicted by the neural network, and determining that a training trigger condition is met when the accuracy is lower than an accuracy threshold or the accuracy decreases by an amplitude threshold;

[0203] When a training trigger condition is met, a round of local training is performed on the neural network according to the historical records stored in the historical database.

[0204] The accuracy can be used as a basis for determining whether the training trigger condition is met, but is not limited to the accuracy in specific implementation.

[0205] For example, the method further includes: monitoring user operations on target content returned based on the suggested parameters; if N consecutive user operations are detected indicating changes to parameter values ​​corresponding to the suggested parameters, it can be considered that the training trigger condition is met.

[0206] Determining the timing of training the neural network model based on the satisfaction of the training trigger conditions can, on the one hand, ensure that the neural network model can give output results with high accuracy, and on the other hand, reduce unnecessary training as much as possible and reduce the power consumption caused by training.

[0207] In some embodiments, the history database includes a first-in-first-out queue, wherein the state information of the terminal device before requesting the target content, the request parameters of the target content, and the state information after the terminal device requests the target content are correspondingly stored in the history database, including:

[0208] When the number of information items in the history database reaches a preset number, the status information of the terminal device before the target content request, the request parameters of the target content, and the status information after the terminal device requests the target content are stored at the end of the first-in-first-out queue.

[0209] The corresponding state information of the terminal device before requesting the target content, the request parameters of the target content, and the state information of the terminal device after requesting the target content are stored as a historical record.

[0210] If the first-in-first-out (FIFO) method is used to store historical records, the overflow problem of the historical database will not occur.

[0211] The preset number of items may be any number. For example, the preset number of items may be 128, 256, or 512.

[0212] In some embodiments, satisfying the preset training trigger condition may further include: the number of updated historical records in the history database reaches a preset ratio, which may be 1 / 3, 1 / 2, 2 / 3, or 1 / 4.

[0213] This is to address the problem of delayed access to target content (images or videos, etc.) on mobile devices due to complex network conditions, unstable network connections, and low mobile device performance. The target content can be: recommended content or pushed content that specifies the content type but not the specific content.

[0214] The disclosed embodiments provide a DQN (Deep Q-Network) energy consumption optimization model. The model considers different mobile devices, different network states, the remaining power of the current device, and the user's active refresh frequency, stores the data obtained from the exploration, and then uses experience replay technology to randomly sample samples to update the model parameters. The data processing is completed on the mobile terminal, and the optimal strategy is calculated from it to dynamically select the best content refresh method, downloaded image format, and preview video format.

[0215] The program flow is as follows:

[0216] S11: The server provides three different sizes of the same resource. For example, images can be available in three resolutions: HD, standard, and low-definition; or videos can be available in three bitrates: 320p, 480p, and 720p. The server can include one or more servers. A server can be a service cluster consisting of multiple servers.

[0217] S12: The mobile terminal collects environmental values ​​based on user actions and trains the DQN model.

[0218] S13: Each time the user takes an action, the trained DQN model is used to obtain optimal request parameters to retrieve images of the specified resolution and videos of the specified bitrate. The optimal request parameters are the aforementioned recommended parameters.

[0219] The technical solution provided by the embodiments of the present disclosure has the following features:

[0220] It has good flexibility and proposes an adaptive content refresh solution that takes into account network status, device status, device power, and user refresh frequency.

[0221] Good user experience. We select the optimal network, image format, and video format based on the current network and device conditions to ensure users receive the best video and image viewing experience.

[0222] Low energy consumption: Reduce device power and time consumption while ensuring user experience.

[0223] Low memory usage. Using the DQN model for calculations can reduce memory requirements.

[0224] The server provides three different sizes of the same resource. For example, images can be available in large, medium, and small resolutions, or videos can be available in three bitrates, for example, 320p, 480p, and 720p.

[0225] When a user requests content, the mobile terminal collects environmental values ​​and trains its own DQN model.

[0226] User actions A may include at least one of the following:

[0227] High definition a1 (High), standard definition a2 (Middle), low definition a3 (Low);

[0228] The environmental values ​​S may each include at least one of the following:

[0229] s1: Network Service, distinguished by the provider name, where WiFi is distinguished by the hotspot name;

[0230] s2: Network Type, which can be WIFI, 5G, 4G, 3G or E;

[0231] s3: Network signal strength, divided into 1-5 strengths;

[0232] s4: mobile phone GPS information, including latitude and longitude;

[0233] s5: The remaining battery information of the mobile phone is recorded in percentage;

[0234] s6: User operation time point, current absolute time.

[0235] The environmental information is a type of the state information of the terminal device, but in a specific implementation, the state information of the terminal device is not limited to the environmental information. For example, the load rate of the terminal device itself may not be included in the environmental value S.

[0236] The reward value R of this algorithm is defined as:

[0237] If the user device takes no more than 300ms to complete the network request, the reward value is 1; otherwise, the reward value is 0.

[0238] The overall algorithm process can be as follows:

[0239] First, initialize Memory D, whose capacity is N;

[0240] Initialize the Q network (the Q network here is the aforementioned neural network model) and randomly generate weights w;

[0241] Initialize the target Q network with weight w′=w;

[0242] Initial state S1;

[0243] Train each step step=1,2,…,I:

[0244] Generate execution action a using ∈-greedy strategy i : Choose a random action with probability ∈, or choose a i =max a Q(S i ,a;w);a i It can be the action to be executed by the terminal device, corresponding to the request parameters of the content request.

[0245] Execute action a i , receive reward i and the new state S i+1 ;.

[0246] The execution actions here include: on-demand or watching high-definition video, on-demand or watching standard-definition video, on-demand or watching low-definition video. This new state means that the terminal device performs action a i After the state.

[0247] Transition samples (S i ,a i ,r i ,S i+1 ) is stored in D; D is the historical database.

[0248] Randomly draw a transitions(S j ,a j ,r j ,S j+1 );

[0249] Calculate the target value for the sample data:

[0250] Use stochastic gradient descent to update the network weights w:

[0251]

[0252] Update the target Q network w′=w every C steps.

[0253] This cycle continues until a good DQN model is trained.

[0254] refer to Figure 5 As shown, the observation value (State, S) is a kind of current state information of the aforementioned terminal device, which is input into the neural network model. The neural network model is as follows Figure 5 The network consists of multiple connected layers. The network outputs a prediction value, which is Q(s, a i , w), indicating the quality score of the state the terminal device will enter after executing the action. Based on a greedy algorithm, for example, the parameters corresponding to the action with the highest quality score can be determined as the recommended parameters. Each time the user list is updated, content is loaded, or more lists are loaded, the collected environment value S is used to retrieve images of a specified resolution and videos of a specified bitrate based on the trained DQN model. This list can be a table in a database storing historical records.

[0255] The DQN model here is a specific example of the aforementioned neural network model.

[0256] Figure 6 A content request method provided by an embodiment of the present disclosure may include:

[0257] Request content;

[0258] The DQN model selects the appropriate bitrate or resolution and requests the video or image.

[0259] Display the requested image or video.

[0260] Collect environmental values ​​and update the DQN model.

[0261] With multiple operators and varying base station standards, mobile devices often encounter unstable networks and slow speeds when accessing network data. This results in content (text, images, videos, etc.) not being displayed to users in a timely, complete, and smooth manner, severely impacting their online experience. At the same time, the quality of content provided by mobile content recommendation platforms is becoming increasingly demanding, placing higher demands on network stability and device performance. However, in some regions, technologies developed based on 4G and 5G networks and high-end mobile devices cannot be directly reused.

[0262] Currently, common solutions for adapting to complex multi-operator networks include supporting media bandwidth resource allocation within operator base stations and proactively adjusting network access or image formats within applications. These approaches have the following drawbacks:

[0263] It is difficult to require different operators in certain regions, such as overseas, to implement corresponding media bandwidth resource allocation adaptation in base stations.

[0264] It only considers the current network status or the current device status, without considering the optimal choice under historical circumstances.

[0265] There are too many network and device states to consider, and it is actually difficult to achieve the processing performance requirements on low-end devices.

[0266] The disclosed embodiment provides a DQL (Deep Q-Network) energy consumption optimization model. This model considers different mobile devices, different network states, the remaining power of the current device, and the user's active refresh frequency. The model stores the data obtained from these explorations, and then uses experience replay technology to randomly sample samples to update model parameters. The data is processed on the mobile terminal, and the optimal strategy is calculated from it. The optimal content refresh method, downloaded image format, and preview video format are dynamically selected. Using local model calculations not only ensures that each person has a unique strategy, but also reduces network request time while ensuring the best user experience, and can reduce the energy consumption of mobile devices.

[0267] To address the issue of delayed access to recommended content (images, videos, etc.) on mobile devices due to complex network conditions, unstable network connections, and low mobile device performance, we proposed a DQN (Deep Q-Network) energy optimization model. This model considers different mobile devices, different network states, the current device's remaining battery life, and the user's active refresh rate. It stores the data obtained from these explorations and then uses experience replay technology to randomly sample samples to update model parameters. Data processing is performed on the mobile device, from which the optimal strategy is calculated to dynamically select the optimal content refresh method, downloaded image format, and preview video format.

[0268] The program process includes at least one of the following:

[0269] The server provides three different sizes of the same resource. For example, images can be in HD, standard, and low-definition resolutions, and videos can be in 320p, 480p, and 720p bitrates.

[0270] When a mobile user requests content (e.g., browsing images or playing videos), the device's environment value S is collected. Before the number of user operations reaches N (128 times), the default strategy is used, and the environment, action, and next state of the operation are recorded. This data is used to train the DQN neural network model.

[0271] When the number of user operations is met, the mobile terminal (i.e., the terminal device) will train the model in the background. The user operation here refers to the user action mentioned above. One user operation corresponds to a content request (or network request).

[0272] After that, each time the user takes an action, the optimal network request parameters are obtained based on the trained DQN model to obtain images of the specified resolution and videos of the specified bitrate.

[0273] The solution provided by the embodiments of the present disclosure is highly flexible and proposes an adaptive content refresh solution that takes into account network status, device battery level, and user refresh frequency, which is also a type of device status information.

[0274] Good user experience. We select the optimal network, image format, and video format based on the current network and device conditions to ensure users receive the best video and image viewing experience.

[0275] Low energy consumption: Reduce device power and time consumption while ensuring user experience.

[0276] It takes up little memory and uses the DQN model for calculation, which can reduce memory requirements.

[0277] The content request method provided by the embodiment of the present disclosure may include:

[0278] The server provides three different sizes of the same resource. For example, images can be available in large, medium, and small resolutions, and videos can be available in 320p, 480p, and 720p bitrates.

[0279] When a mobile user requests content (i.e., browses pictures, plays videos), the environment value S on the device side is collected.

[0280] When the number of user operations does not reach N (128 times), the default strategy is used and the environment, action, and next state of the operation are recorded. This data will be used to train the DQN neural network model.

[0281] The default policy is: When the environment information obtained is a WiFi network, the remaining battery is greater than 80%, and the network signal strength is greater than or equal to 4, the default selection is A2 standard definition image and 480p bitrate video. In other cases, the default selection is A3 low definition image and 320p bitrate video.

[0282] The logic for the next state is that if the image loading time is less than 300ms or the video cache progress within 300ms is greater than 6s, the state is upgraded to the next level (for example: A3 -> A2);

[0283] When the image loading time is greater than 1000s or the video does not start playing within 1000s, the status is downgraded by one level (for example: A2->A3); in other cases, it remains unchanged.

[0284] When the number of user operations is met, the mobile terminal will train the model in the background,

[0285] User action A includes one of the following:

[0286] High definition a1 (High), standard definition a2 (Middle), low definition a3 (Low);

[0287] 3.2 The environmental value S is at least one of the following:

[0288] s1: Network Service: Distinguished by the provider name, with WiFi distinguished by the hotspot name;

[0289] s2: Network Type: divided into WIFI, 5G, 4G, 3G, E;

[0290] s3: Network signal strength: divided into 1-5 strengths;

[0291] s4: mobile phone GPS information: including latitude and longitude;

[0292] s5: Mobile phone remaining power information: recorded in percentage;

[0293] s6: User operation time point: current absolute time value (from 0:00 to 24:00);

[0294] Determine the reward value R. The R value is expressed as the target (Traget). The Q network is a matrix simulated by a neural network. Since there are too many attributes to maintain in the mobile terminal memory, a neural network is chosen to simulate it, as shown in the following example:

[0295] Q-table Low-definition A1 SD A2 HD A3 S1(s1,s2,…,s6) Q(1，1) Q(1，2) Q(1，3) S2(s1,s2,…,s6) Q(2，1) Q(2，2) Q(2，3) ……

[0296] Initialize the matrix W (all to 0), randomly select the action for the first time using the greedy strategy. If you choose a2 and then select state S2, you get a reward of 1, then Q(1,2) = 1

[0297] refer to Figure 7 As shown, it may include:

[0298] First, initialize the database D, whose capacity is N (128);

[0299] Initialize the Q network and randomly generate weights w;

[0300] Initialize the target Q network with weight w′=w;

[0301] Initialization (initial state) S1;

[0302] Training each step step=1,2,…,I:

[0303] Generate action a using reward greedy (∈-greedy)∈-greedy strategy i : Choose a random action with probability ∈, or choose a i =max a Q(S i ,a;w);

[0304] Execute action a i , receive reward i and the new state S i+1 ;

[0305] The transition sample (S i ,a i ,r i ,S i+1 ) is stored in D;

[0306] Randomly draw a transitions(S j ,a j ,r j ,S j+1 );

[0307] Calculate the target value for the sample data:

[0308] Use stochastic gradient descent to update the network weights w:

[0309]

[0310] Update the target Q network w′=w every C steps.

[0311] This cycle continues until a good DQN model is trained.

[0312] The DQN model learns from historical records obtained from the historical database (D). After learning, it generates predicted values ​​and actual Q values, calculates loss values, and updates the DQN model's network parameters (e.g., weights and / or biases) based on the direction propagation of the loss values, thereby achieving local training of the DQN model. After this, each time the terminal device takes an action (loads an image or video), it displays images of a specified resolution and plays videos of a specified bitrate based on the environmental value S required in 3.2, based on the trained DQN model.

[0313] refer to Figure 8 As shown, the content request method provided by the embodiment of the present disclosure may include:

[0314] Display pictures or play videos;

[0315] Collect environmental value S;

[0316] Determine if a DQN model exists;

[0317] If it exists, generate Q(S, a) based on the DQN model i , W);

[0318] Based on the Q(S, a) generated by the DQN model i , W) select bit rate and resolution;

[0319] If the DQN model does not exist, determine whether the database D has N values, where N values ​​are the predetermined number of historical records.

[0320] If not, generate Q(S, a) based on the default strategy i ,,W);

[0321] If so, train the DQN model;

[0322] Store Q(S, a i , W) to database D;

[0323] Display videos with a specified bitrate or resolution.

[0324] like Figure 9 As shown, the embodiment of the present disclosure provides a content request device, including:

[0325] A first determining module 110 is configured to determine current status information of a terminal device;

[0326] A parameter module 120 is configured to input the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting a target content from a content server;

[0327] The request module 130 is configured to send a content request carrying the suggested parameters to the content server.

[0328] In some embodiments, the first determination module 110 , the parameter module 120 and the request module 130 may all be program modules; after being executed by a processor, the program modules may implement the functions of the above modules.

[0329] In some embodiments, the first determination module 110, the parameter module 120, and the request module 130 may be a combination of software and hardware modules, including but not limited to various programmable arrays, including but not limited to field programmable arrays and / or complex programmable arrays.

[0330] In some other embodiments, the first determination module 110 , the parameter module 120 , and the request module 130 may be pure hardware modules; the pure hardware modules include but are not limited to application-specific integrated circuits.

[0331] In some embodiments, the parameter module 120 is specifically used to input the current state information and the nth set of alternative parameters into the neural network model to obtain predicted state information; wherein, the predicted state information indicates: the state that the terminal device is ready to enter when requesting the target content with the nth alternative parameter; n is a positive integer less than N; N is the total number of sets of alternative parameters; a set of the alternative parameters includes at least one request parameter; when the state that the terminal device is ready to enter meets the preset conditions, the nth set of alternative parameters is used as the recommended parameters.

[0332] In some embodiments, the apparatus further comprises:

[0333] a calculation module, configured to calculate a quality score value based on the state of the terminal device to be entered and the nth set of candidate parameters;

[0334] The determination module is configured to determine the terminal ready-to-enter state with the largest quality score as the terminal ready-to-enter state that meets the preset condition.

[0335] In some embodiments, the current status information includes at least one of the following:

[0336] Network status information of the terminal device;

[0337] Remaining power information of the terminal device;

[0338] location information of the terminal device;

[0339] time information when the terminal device requests the target content;

[0340] The terminal device currently outputs content parameters of the target content.

[0341] In some embodiments, the network status information indicates at least one of the following:

[0342] The network provider to which the terminal device is currently connected;

[0343] The type of network to which the terminal device is currently connected;

[0344] The current network signal strength of the terminal device;

[0345] The current network bandwidth of the terminal device.

[0346] In some embodiments, the apparatus further comprises:

[0347] A second determination module is configured to determine whether the terminal device stores model information of the neural network model that has completed at least one round of local training;

[0348] The parameter module 120 is specifically used to run the neural network model according to the model information when the terminal device stores the model information of the neural network model that has completed at least one round of local training; input the current state information into the neural network model running locally on the terminal device to obtain the recommended parameters.

[0349] In some embodiments, the parameter module 120 is further specifically used to determine the recommended parameters based on the default strategy and the current status information when the terminal device does not store model information of the neural network model that has completed at least one round of local training.

[0350] In some embodiments, the parameter module 120 is specifically configured to determine the mth set of alternative parameters as the recommended parameters when the network type connected to the terminal device is a non-traffic sensitive network, the remaining power of the terminal device is greater than a power threshold, and / or the network signal strength connected to the terminal device is greater than a strength threshold; otherwise, determine the m+1th set of alternative parameters as the recommended parameters;

[0351] The content quality of the target content output by the terminal device using the mth set of candidate parameters is higher than the content quality of the target content output by the terminal device using the (m+1)th set of candidate parameters.

[0352] In some embodiments, the alternative parameters also include: the m+2th set of alternative parameters, wherein the content quality when the terminal device outputs the target content with the m+2th set of alternative parameters is higher than the content quality when the terminal device outputs the target content with the m+1th set of alternative parameters.

[0353] In some embodiments, the apparatus further comprises:

[0354] a storage module, configured to store the state information of the terminal device before the target content is requested, the request parameters of the target content, and the state information after the terminal device requests the target content, in a history database;

[0355] The training module is used to perform a round of local training on the neural network model using the historical data recorded in the historical database as sample data when the number of information items in the historical database reaches a preset number.

[0356] In some embodiments, the historical database includes: a first-in-first-out queue, wherein the storage module is specifically used to store the status information of the terminal device before the target content request, the request parameters of the target content, and the status information after the terminal device requests the target content at the end of the first-in-first-out queue when the number of information items in the historical database reaches a preset number.

[0357] An embodiment of the present disclosure provides an electronic device, including:

[0358] a memory for storing processor-executable instructions;

[0359] Processor, respectively memory connected;

[0360] The processor is configured to execute the content request method provided by any of the aforementioned technical solutions by executing computer-executable instructions stored in the memory.

[0361] The electronic device may be any of the aforementioned terminal devices.

[0362] The processor may include various types of storage media, which are non-transitory computer storage media that can continue to store information after the mobile terminal loses power.

[0363] The processor can be connected to the memory via a bus or the like, and is used to read the executable program stored in the memory, for example, Figures 1 to 8 At least one of any of the methods shown.

[0364] Figure 10 FIG. 8 is a block diagram of a terminal device 800 according to an exemplary embodiment. For example, the terminal device 800 may be a mobile phone, a mobile computer, or the like.

[0365] Reference Figure 10 The terminal device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power supply component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0366] The processing component 802 generally controls the overall operation of the terminal device 800, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

[0367] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the terminal device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0368] The power component 806 provides power to the various components of the terminal device 800. The power component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the terminal device 800.

[0369] The multimedia component 808 includes a screen that provides an output interface between the terminal device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor can not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating state, such as a shooting state or a video state, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have a focal length and optical zoom capability.

[0370] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the terminal device 800 is in an operating state, such as a call state, a recording state, and a voice recognition state, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0371] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include but are not limited to: a home button, volume buttons, a start button, and a lock button.

[0372] Sensor assembly 814 includes one or more sensors for providing status assessment of various aspects for terminal device 800. For example, sensor assembly 814 can detect the open / closed state of device 800, the relative positioning of components, such as the display and keypad of terminal device 800, and sensor assembly 814 can also detect the position change of terminal device 800 or a component of terminal device 800, the presence or absence of user contact with terminal device 800, the orientation or acceleration / deceleration of terminal device 800 and the temperature change of terminal device 800. Sensor assembly 814 can include proximity sensor, is configured to detect the presence of nearby objects when there is no physical contact. Sensor assembly 814 can also include light sensor, such as CMOS or CCD image sensor, for use in imaging applications. In some embodiments, this sensor assembly 814 can also include acceleration sensor, gyroscope sensor, magnetic sensor, pressure sensor or temperature sensor.

[0373] The communication component 816 is configured to facilitate wired or wireless communication between the terminal device 800 and other devices. The terminal device 800 can access a wireless network based on a communication standard, such as Wi-Fi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0374] In an exemplary embodiment, the terminal device 800 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0375] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by the processor 820 of the terminal device 800 to perform the above method. For example, the non-transitory 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.

[0376] The embodiment of the present disclosure provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the mobile terminal, the mobile terminal can execute the content request method provided by any of the above embodiments, and can execute the following steps: Figures 1 to 3 At least one of any of the methods shown.

[0377] Other embodiments of the present invention will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the invention being indicated by the following claims.

[0378] It should be understood that the present invention is not limited to the exact construction 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 invention is limited only by the appended claims.

Claims

1. A content request method, characterized in that: Applied to terminal equipment, including: Determining current status information of the terminal device; Inputting the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting target content from a content server; Sending a content request carrying the suggested parameters to the content server; The step of inputting the current state information into a neural network model running in the terminal device includes: Inputting the current state information and the nth set of candidate parameters into the neural network model to obtain predicted state information; wherein the predicted state information indicates: a state that the terminal device is ready to enter when requesting target content with the nth candidate parameter; n is a positive integer less than N; N is the total number of sets of candidate parameters; and a set of candidate parameters includes at least one request parameter; When the state to be entered by the terminal device satisfies a preset condition, taking the nth set of candidate parameters as the suggested parameters; The method further includes: calculating a quality score value based on the terminal device's ready-to-enter state and the nth set of candidate parameters; and determining the terminal device's ready-to-enter state with the largest quality score value as the terminal device's ready-to-enter state that meets the preset condition.

2. The method according to claim 1, characterized in that The current status information includes at least one of the following: Network status information of the terminal device; Remaining power information of the terminal device; location information of the terminal device; time information when the terminal device requests the target content; The terminal device currently outputs content parameters of the target content.

3. The method according to claim 2, characterized in that The network status information indicates at least one of the following: The network provider to which the terminal device is currently connected; The type of network to which the terminal device is currently connected; The current network signal strength of the terminal device; The current network bandwidth of the terminal device.

4. The method according to claim 1, wherein The method further comprises: Determining whether the terminal device stores model information of the neural network model that has completed at least one round of local training; The step of inputting the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting a target content from a content server includes: When the terminal device stores model information of the neural network model that has completed at least one round of local training, running the neural network model according to the model information; The current state information is input into the neural network model running locally on the terminal device to obtain the recommended parameters.

5. The method according to claim 4, characterized in that The method further comprises: When the terminal device does not store the model information of the neural network model that has completed at least one round of local training, the recommended parameters are determined according to the default strategy and the current state information.

6. The method according to claim 5, characterized in that The determining the recommended parameters according to the default policy and the current state information includes: When the network type connected to the terminal device is a non-traffic sensitive network, the remaining power of the terminal device is greater than a power threshold, and / or the network signal strength connected to the terminal device is greater than a strength threshold, determining the mth set of alternative parameters as the recommended parameters; otherwise, determining the m+1th set of alternative parameters as the recommended parameters; The content quality of the target content output by the terminal device using the mth set of candidate parameters is higher than the content quality of the target content output by the terminal device using the (m+1)th set of candidate parameters.

7. The method according to claim 6, characterized in that The alternative parameters also include: an m+2th set of alternative parameters, wherein the content quality of the target content output by the terminal device using the m+2th set of alternative parameters is higher than the content quality of the target content output by the terminal device using the m+1th set of alternative parameters.

8. The method according to claim 5, characterized in that The method further comprises: storing the state information of the terminal device before requesting the target content, the request parameters of the target content, and the state information of the terminal device after requesting the target content in a history database; When the number of information items in the historical database reaches a preset number, the neural network model is subjected to a round of local training using the historical data recorded in the historical database as sample data.

9. The method according to claim 8, characterized in that The historical database includes a first-in-first-out queue, wherein the state information of the terminal device before the target content is requested, the request parameters of the target content, and the state information after the terminal device requests the target content are correspondingly stored in the historical database, including: When the number of information items in the history database reaches a preset number, the status information of the terminal device before the target content request, the request parameters of the target content, and the status information after the terminal device requests the target content are stored at the end of the first-in-first-out queue.

10. A content request device, characterized in that: Applied to terminal equipment, including: A first determining module, configured to determine current status information of the terminal device; a parameter module, configured to input the current state information into a neural network model running in the terminal device to obtain recommended parameters for requesting target content from a content server; A request module, configured to send a content request carrying the suggested parameters to the content server; The step of inputting the current state information into a neural network model running in the terminal device includes: Inputting the current state information and the nth set of candidate parameters into the neural network model to obtain predicted state information; wherein the predicted state information indicates: a state that the terminal device is ready to enter when requesting target content with the nth candidate parameter; n is a positive integer less than N; N is the total number of sets of candidate parameters; and a set of candidate parameters includes at least one request parameter; When the state to be entered by the terminal device satisfies a preset condition, taking the nth set of candidate parameters as the suggested parameters; The device further comprises: a calculation module, configured to calculate a quality score value based on the state of the terminal device to be entered and the nth set of candidate parameters; The determination module is configured to determine the terminal ready-to-enter state with the largest quality score as the terminal ready-to-enter state that meets the preset condition.

11. The device according to claim 10, characterized in that The current status information includes at least one of the following: Network status information of the terminal device; Remaining power information of the terminal device; location information of the terminal device; time information when the terminal device requests the target content; The terminal device currently outputs content parameters of the target content.

12. The device according to claim 11, characterized in that The network status information indicates at least one of the following: The network provider to which the terminal device is currently connected; The type of network to which the terminal device is currently connected; The current network signal strength of the terminal device; The current network bandwidth of the terminal device.

13. The device according to claim 10, characterized in that The device further comprises: A second determination module is configured to determine whether the terminal device stores model information of the neural network model that has completed at least one round of local training; The parameter module is specifically used to run the neural network model according to the model information when the terminal device stores the model information of the neural network model that has completed at least one round of local training; input the current state information into the neural network model running locally on the terminal device to obtain the recommended parameters.

14. The device according to claim 13, characterized in that The parameter module is further specifically used to determine the recommended parameters based on the default strategy and the current state information when the terminal device does not store the model information of the neural network model that has completed at least one round of local training.

15. The device according to claim 14, characterized in that The parameter module is further specifically configured to determine the mth set of alternative parameters as the recommended parameters when the network type connected to the terminal device is a non-traffic sensitive network, the remaining power of the terminal device is greater than a power threshold, and / or the network signal strength connected to the terminal device is greater than a strength threshold; otherwise, determine the m+1th set of alternative parameters as the recommended parameters; The content quality of the target content output by the terminal device using the mth set of candidate parameters is higher than the content quality of the target content output by the terminal device using the (m+1)th set of candidate parameters.

16. The device according to claim 15, characterized in that The alternative parameters also include: an m+2th set of alternative parameters, wherein the content quality of the target content output by the terminal device using the m+2th set of alternative parameters is higher than the content quality of the target content output by the terminal device using the m+1th set of alternative parameters.

17. The device according to claim 15, characterized in that The device further comprises: a storage module, configured to store the state information of the terminal device before the target content is requested, the request parameters of the target content, and the state information after the terminal device requests the target content, in a history database; The training module is used to perform a round of local training on the neural network model using the historical data recorded in the historical database as sample data when the number of information items in the historical database reaches a preset number.

18. The device according to claim 17, characterized in that The historical database includes: a first-in-first-out queue, wherein the storage module is specifically used to store the status information of the terminal device before the target content request, the request parameters of the target content, and the status information after the terminal device requests the target content at the end of the first-in-first-out queue when the number of information items in the historical database reaches a preset number.

19. An electronic device, characterized in that: include: a memory for storing processor-executable instructions; a processor connected to the memory; The processor is configured to execute the method for providing content request according to any one of claims 1 to 9.

20. A non-transitory computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, implement the content request method provided in any one of claims 1 to 9.

Images