Information recommendation method, vehicle and storage medium

Abstract

This invention discloses an information recommendation method, vehicle, and storage medium, relating to the field of information recommendation technology. The method enables a vehicle-machine interaction system to quickly and proactively recommend information to users. The specific solution includes: acquiring collected information from each of multiple users in the vehicle; obtaining characteristic data for each user based on the collected information; acquiring target voice data in response to a wake-up operation; determining target information based on the target voice data and the characteristic data for each of the multiple users; and pushing the target information to the users in the vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of information recommendation, and in particular to an information recommendation method, a vehicle, and a storage medium. Background Art

[0002] Intelligent human-computer voice interaction enables efficient information exchange between humans and computer devices using natural language. Current intelligent human-computer voice interaction typically involves multiple rounds of dialogue. This involves performing voice recognition on the user's voice information, then performing entity recognition, and then filling in the slots with the entities. Once the slots are filled, feedback is provided to the user.

[0003] The vehicle-to-machine interaction system uses intelligent human-machine voice interaction technology. It needs to engage in multiple rounds of dialogue with the driver before it can understand the driver's true intentions and recommend information to the driver. However, this prevents the system from quickly and proactively recommending information to the driver, distracting the driver and affecting driving safety. Summary of the Invention

[0004] The present invention provides an information recommendation method, device, vehicle and storage medium, which can enable a vehicle-machine interaction system to quickly and actively recommend information to users.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] In a first aspect, the present invention provides an information recommendation method, the method comprising:

[0007] Obtaining collected information of each of a plurality of users in the vehicle;

[0008] Obtain characteristic data of each user based on the collected information of each user;

[0009] In response to the wake-up operation, acquiring target voice data;

[0010] determining target information based on target voice data and feature data of each of the plurality of users;

[0011] Push targeted information to users in the vehicle.

[0012] By adopting the information recommendation method of the present invention, during the driving process of the vehicle, the vehicle terminal can determine the characteristic data of each user based on the collected information of each user in the vehicle. When the user wakes up the vehicle-machine interaction system of the vehicle terminal and the vehicle terminal obtains the target voice data of the user in the vehicle, the vehicle terminal will determine the target information that meets the user intention corresponding to the target voice data and the interests of each user in the vehicle based on the target voice data and the characteristic data of each user in the vehicle, and push the target information to the user in the vehicle. Compared with the prior art, the user in the vehicle needs to go through multiple rounds of dialogue with the vehicle-machine interaction system before obtaining the target information that meets the user's intention and interests. The information recommendation method provided by the embodiment of the present invention can quickly and proactively recommend information to the users in the vehicle. Moreover, in the process of recommending information, the embodiment of the present invention takes into account the characteristic data of each user in the vehicle, so that the recommended information can meet the interests of all users in the vehicle, which makes the recommendation results more comprehensive.

[0013] In one possible implementation, determining the target information based on the target voice data and the feature data of each of the multiple users includes:

[0014] Determine user intent based on target voice data;

[0015] Determine target information based on user intent and each user's characteristic data.

[0016] In one possible implementation, determining target information based on user intent and characteristic data of each user includes:

[0017] Determine the target category corresponding to the user’s intent;

[0018] Obtain multiple candidate information according to the target category;

[0019] determining, based on characteristic data of the first user corresponding to the target voice data, a first coefficient for each piece of candidate information corresponding to the first user, the first coefficient being used to indicate a degree of interest of the first user in the candidate information;

[0020] Determining, based on characteristic data of each second user and a pre-stored recommendation model, a second coefficient for each piece of candidate information corresponding to each second user, the second coefficient being used to indicate a degree of interest of the second user in the candidate information, where the second user is a user other than the first user among the multiple users;

[0021] Target information is determined from a plurality of candidate information according to the determined first coefficient and second coefficient.

[0022] In a possible implementation, determining the second coefficient of each piece of candidate information corresponding to each second user based on the feature data of each second user and a pre-stored recommendation model includes:

[0023] If the collected information of the second user meets the preset conditions, determining a second coefficient of each candidate information corresponding to the second user according to the characteristic data of the second user;

[0024] If the collected information of the second user does not meet the preset condition, a preset value is determined as the second coefficient of each candidate information corresponding to the second user.

[0025] In a possible implementation, determining the first coefficient of each piece of candidate information corresponding to the first user based on the feature data of the first user corresponding to the target voice data includes:

[0026] Obtain the weight of each candidate information in each attribute of the target category;

[0027] Determining a recommendation coefficient for each attribute of the target category based on the feature data of the first user corresponding to the target voice data, wherein the recommendation coefficient for each attribute is used to indicate the first user's interest level in the attribute;

[0028] The first coefficient of each candidate information is determined according to the weight of each candidate information in each attribute and the recommendation coefficient corresponding to each attribute.

[0029] In a possible implementation, the collected information includes a captured image, the captured image includes target portrait information of a target user, and the target user is any one of the multiple users;

[0030] The above method obtains characteristic data of each user based on the collected information of each user, including:

[0031] If there is historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user meets the preset conditions, and characteristic data of the target user is determined based on the target portrait information and the matching historical portrait information;

[0032] If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user does not meet the preset conditions, and characteristic data of the target user is obtained based on the target portrait information.

[0033] In one possible implementation, the above-mentioned acquisition of target user feature data based on the target portrait information and the matching historical portrait information includes:

[0034] Determine current feature data based on target portrait information;

[0035] Update the user portrait corresponding to the historical portrait information based on the current feature data;

[0036] Determine the characteristic data of target users based on the updated user portrait.

[0037] In a possible implementation, the collected information includes collected voice data, and the collected voice data includes recording data of a target user, where the target user is any one of the multiple users;

[0038] The above method obtains characteristic data of each user based on the collected information of each user, including:

[0039] If there is historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user meets the preset conditions, and characteristic data of the target user is obtained based on the collected voice data and the matching historical voice data;

[0040] If there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user does not meet the preset conditions, and characteristic data of the target user is obtained based on the collected voice data.

[0041] In one possible implementation, the step of obtaining the collected information of each of the multiple users in the vehicle includes:

[0042] The collected information of the user of each seat is obtained through the collection device corresponding to each seat in the vehicle, and the collected information of each user among the multiple users is obtained.

[0043] In a possible implementation, the collected information includes captured images and collected voice data, the captured images include target portrait information of a target user, and the collected voice data includes recorded data of the target user, where the target user is any one of the multiple users;

[0044] The information recommendation method further includes: sending collected information to a server, the collected information being used by the server to determine feature data;

[0045] The above method obtains characteristic data of each user based on the collected information of each user, including:

[0046] In response to the wake-up operation, if it is determined that no feature data sent by the server has been received, if there is historical portrait information matching the target portrait information in the pre-stored portrait information, and there is historical voice data matching the collected voice data in the pre-stored voice data, then determining that the collected information of the target user meets the preset conditions, and obtaining the feature data of the target user based on the user portrait corresponding to the historical portrait information and the historical voice data;

[0047] If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, or there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user does not meet the preset conditions, and the preset data is used as the feature data of the target user.

[0048] In a second aspect, the present invention provides an information recommendation device, comprising:

[0049] an acquiring unit, configured to acquire collected information of each of a plurality of users in the vehicle;

[0050] A determination unit, configured to obtain characteristic data of each user based on the collected information of each user;

[0051] The acquisition unit is further configured to acquire target voice data in response to a wake-up operation;

[0052] The determination unit is further configured to determine target information based on the target voice data and feature data of each of the multiple users;

[0053] The sending unit is used to push target information to users in the vehicle.

[0054] In a possible implementation, the determining unit is specifically configured to:

[0055] Determine user intent based on target voice data;

[0056] Determine target information based on user intent and each user's characteristic data.

[0057] In a possible implementation, the determining unit is specifically configured to:

[0058] Determine the target category corresponding to the user’s intent;

[0059] Obtain multiple candidate information based on target category and user intent;

[0060] determining, based on characteristic data of the first user corresponding to the target voice data, a first coefficient for each piece of candidate information corresponding to the first user, the first coefficient being used to indicate a degree of interest of the first user in the candidate information;

[0061] Determining, based on characteristic data of each second user and a pre-stored recommendation model, a second coefficient for each piece of candidate information corresponding to each second user, the second coefficient being used to indicate a degree of interest of the second user in the candidate information, where the second user is a user other than the first user among the multiple users;

[0062] Target information is determined from a plurality of candidate information according to the determined first coefficient and second coefficient.

[0063] In a possible implementation, the determining unit is specifically configured to:

[0064] If the collected information of the second user meets the preset conditions, determining a second coefficient of each candidate information corresponding to the second user according to the characteristic data of the second user;

[0065] If the collected information of the second user does not meet the preset condition, a preset value is determined as the second coefficient of each candidate information corresponding to the second user.

[0066] In a possible implementation, the determining unit is specifically configured to:

[0067] Obtain the weight of each candidate information in each attribute of the target category;

[0068] Determining a recommendation coefficient for each attribute of the target category based on the feature data of the first user corresponding to the target voice data, where the recommendation coefficient for each attribute is used to indicate the first user's interest in the attribute;

[0069] The first coefficient of each candidate information is determined according to the weight of each candidate information in each attribute and the recommendation coefficient corresponding to each attribute.

[0070] In a possible implementation, the collected information includes a captured image, the captured image includes target portrait information of a target user, and the target user is any one of the multiple users;

[0071] The above-mentioned determination unit is specifically used to:

[0072] If there is historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user meets the preset conditions, and characteristic data of the target user is obtained based on the target portrait information and the matching historical portrait information;

[0073] If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user does not meet the preset conditions, and characteristic data of the target user is obtained based on the target portrait information.

[0074] In a possible implementation, the determining unit is specifically configured to:

[0075] Determine current feature data based on target portrait information;

[0076] Update the user portrait corresponding to the historical portrait information based on the current feature data;

[0077] Determine the characteristic data of target users based on the updated user portrait.

[0078] In a possible implementation, the collected information includes collected voice data, and the collected voice data includes recording data of a target user, where the target user is any one of the multiple users;

[0079] The above-mentioned determination unit is specifically used to:

[0080] If there is historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user meets the preset conditions, and characteristic data of the target user is obtained based on the collected voice data and the matching historical voice data;

[0081] If there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user meets the non-preset condition, and characteristic data of the target user is obtained based on the collected voice data.

[0082] In a possible implementation, the acquisition unit is specifically configured to:

[0083] The collected information of the user of each seat is obtained through the collection device corresponding to each seat in the vehicle, and the collected information of each user among the multiple users is obtained.

[0084] In a possible implementation, the sending unit is further configured to send collected information to the server, where the collected information is used by the server to determine feature data.

[0085] The above-mentioned determination unit is specifically used to:

[0086] In response to the wake-up operation, if it is determined that no feature data sent by the server has been received, if there is historical portrait information matching the target portrait information in the pre-stored portrait information, and there is historical voice data matching the collected voice data in the pre-stored voice data, then determining that the collected information of the target user meets the preset conditions, and obtaining the feature data of the target user based on the user portrait corresponding to the historical portrait information and the historical voice data;

[0087] If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, or there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user does not meet the preset conditions, and the preset data is used as the feature data of the target user.

[0088] In a third aspect, the present invention provides a vehicle comprising: a processor and a memory. The memory is configured to store computer program code, the computer program code comprising computer instructions. When the processor executes the computer instructions, the vehicle performs the information recommendation method of the first aspect and any possible implementation thereof.

[0089] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon. When the computer instructions are executed on a vehicle, the vehicle executes an information recommendation method as in the first aspect or any one of the possible implementations of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0090] Figure 1 A schematic diagram of the structure of an information recommendation system provided by an embodiment of the present invention;

[0091] Figure 2 A schematic structural diagram of a vehicle provided in an embodiment of the present invention;

[0092] Figure 3 This is a flow chart of an information recommendation method according to an embodiment of the present invention;

[0093] Figure 4 This is a second flow chart of the information recommendation method provided by an embodiment of the present invention;

[0094] Figure 5 This is a third flow chart of the information recommendation method provided by an embodiment of the present invention;

[0095] Figure 6 This is a fourth flow chart of the information recommendation method provided by an embodiment of the present invention;

[0096] Figure 7 This is a structural diagram of an information recommendation device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0097] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0098] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present disclosure, unless otherwise specified, "multiple" means two or more. In addition, the use of "based on" or "according to" means openness and inclusiveness, because the process, steps, calculations or other actions "based on" or "according to" one or more of the conditions or values ​​may be based on additional conditions or values ​​beyond the stated in practice.

[0099] In order to enable the vehicle-machine interaction system to quickly and proactively recommend information to users, an embodiment of the present invention provides an information recommendation method, a vehicle, and a storage medium. The vehicle-mounted terminal obtains the collected information of each of the multiple users in the vehicle in real time through the collection device, and obtains the characteristic data of each user based on the collected information of each user. In addition, the vehicle-mounted terminal also responds to the wake-up operation of the user in the vehicle, obtains the target voice data through the collection device, and determines the target information based on the target voice data and the characteristic data of each of the multiple users, and finally pushes the target information to the user in the vehicle. In this way, during the process of interaction between the user in the vehicle and the vehicle-machine interaction system of the vehicle-mounted terminal, when the vehicle obtains the user's target voice information, the vehicle-machine interaction system of the vehicle-mounted terminal can quickly and proactively recommend to the user target information that meets the user's intention and the preferences of all users in the vehicle.

[0100] The information recommendation method provided in this embodiment of the present invention is implemented by an information recommendation device. The information recommendation device can be a vehicle, an onboard terminal within a vehicle, a central processing unit (CPU) within the vehicle, or a client within the vehicle for performing information recommendation. This embodiment of the present invention uses the onboard terminal as an example to illustrate the information recommendation method provided by this application.

[0101] In one scenario, when executing the information recommendation method of an embodiment of the present invention, an in-vehicle terminal, after acquiring collected information about each of multiple users in the vehicle, can perform facial recognition, voiceprint recognition, and feature data extraction on the collected information. Furthermore, after acquiring target voice data, the in-vehicle terminal determines target information based on the target voice data and the feature data of each of the multiple users, and pushes the target information to the users in the vehicle.

[0102] In another scenario, the above feature data extraction process can be performed by a server.

[0103] Specifically, the information recommendation method provided by the embodiment of the present invention may be applicable to an information recommendation system. Figure 1 A structural diagram of the information recommendation system is shown in FIG. Figure 1 As shown, the information recommendation system may include a vehicle 11, a first server 12, and a second server 13. The vehicle is equipped with an onboard terminal and a data collection device. The onboard terminal and the data collection device are connected to the first server 12 and the second server 13 respectively via wired or wireless communication.

[0104] The first server 12 is configured to receive collected information of each user among a plurality of users in the vehicle, determine feature data of each user based on the collected information, and send the feature data of each user to the vehicle-mounted terminal.

[0105] The second server 13 is configured to determine a plurality of candidate information according to the user's intention and send the plurality of candidate information to the vehicle-mounted terminal.

[0106] The collection device is used to collect information from each of the multiple users in the vehicle in real time and send the collected information to the vehicle terminal. It is also used to respond to the user's wake-up operation, collect the user's target voice information, and send the target voice information to the vehicle terminal. The collection device is installed on each seat in the vehicle.

[0107] Exemplarily, the acquisition device may include an image acquisition device and an audio acquisition device, wherein the image acquisition device is used to acquire the user's portrait information, and the audio acquisition device is used to acquire the user's voice data. For example, the image acquisition device is a camera, and the audio acquisition device is a microphone.

[0108] The vehicle-mounted terminal is used to send the collected information obtained by the collection device to the first server 12, and receive the characteristic data of each user sent by the first server 12; it is also used to determine the user intention based on the target voice data obtained by the collection device, send the user intention to the second server 13, and receive multiple candidate information sent by the second server 13; it is also used to determine the target information based on the multiple candidate information and the characteristic data of each user in the multiple users, and push the target information to the users in the vehicle.

[0109] Figure 2 A schematic diagram of the vehicle structure is shown in Figure 2. Figure 2 As shown, the vehicle may include: a processor 21 , a memory 22 , a communication interface 23 and a bus 24 . The processor 21 , the memory 22 and the communication interface 23 may be connected via the communication bus 24 .

[0110] The processor 21 is the control center of the vehicle and can be a single processor 21 or a collective term for multiple processing elements. For example, the processor 21 can be a general-purpose CPU or other general-purpose processor 21. The general-purpose processor 21 can be a microprocessor 21 or any conventional processor 21.

[0111] As an embodiment, the processor 21 may include one or more CPUs, for example, Figure 2 CPU0 and CPU1 are shown.

[0112] The memory 22 may be a read-only memory 22 (ROM) or other type of static storage device that can store static information and instructions, a random access memory 22 (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory 22 (EEPROM), a disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0113] In one possible implementation, memory 22 may exist independently of processor 21 and may be connected to processor 21 via bus 24 to store instructions or program code. When processor 21 calls and executes the instructions or program code stored in memory 22, the information recommendation method provided in the following embodiments of the present invention can be implemented.

[0114] In another possible implementation, the memory 22 may also be integrated with the processor 21 .

[0115] The communication interface 23 is used to connect the vehicle to other devices via a communication network, such as Ethernet, a radio access network (RAN), a wireless local area network (WLAN), etc. The communication interface 23 may include a receiving unit for receiving data and a transmitting unit for transmitting data.

[0116] The bus 24 may be an Industry Standard Architecture (ISA) bus 24, a Peripheral Component Interconnect (PCI) bus 24, or an Extended Industry Standard Architecture (EISA) bus 24. The bus 24 may be divided into an address bus 24, a data bus 24, a control bus 24, and the like. For ease of representation, Figure 2 Only one thick line is used in the figure, but this does not mean that there is only one bus 24 or only one type of bus 24.

[0117] It should be pointed out that Figure 2 The structure shown in the figure does not constitute a limitation on the vehicle, except Figure 2In addition to the components shown, the vehicle may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0118] The information recommendation method provided by the embodiment of the present invention is described below with reference to the accompanying drawings.

[0119] like Figure 3 As shown, the information recommendation method provided by the embodiment of the present invention includes the following steps 301 to 305.

[0120] 301. The vehicle-mounted terminal obtains collected information of each of multiple users in the vehicle.

[0121] Optionally, a collection device is installed at a corresponding position of each seat in the vehicle for real-time collection of information within the vehicle. The vehicle terminal obtains the collected information of the user in each seat through the collection device corresponding to each seat in the vehicle, thereby obtaining the collected information of each user among the multiple users.

[0122] In one scenario, when multiple users enter the vehicle, the collection device on the corresponding seat begins to collect the collection information of the user in the seat, and sends all the collected collection information to the vehicle terminal, so that the vehicle terminal obtains the collection information of each user in the vehicle.

[0123] 302. The vehicle-mounted terminal obtains characteristic data of each user based on the collected information of each user.

[0124] 303. The vehicle-mounted terminal obtains target voice data in response to the wake-up operation.

[0125] In one scenario, when any of the multiple users in a vehicle wants to interact with the vehicle-mounted terminal's vehicle-machine interaction system, they first need to wake up the vehicle-mounted terminal's vehicle-machine interaction system. After the vehicle-mounted terminal's vehicle-machine interaction system is woken up, any of the multiple users in the vehicle can interact with the vehicle-mounted terminal's vehicle-machine interaction system through voice. At this point, the collection device can collect the user's voice recording data and send it to the vehicle-mounted terminal. This voice recording data is the target voice data.

[0126] Optionally, when waking up the vehicle-mounted terminal's vehicle-machine interaction system, the user can do so by clicking a wake-up button on the vehicle-mounted terminal's vehicle-machine interaction system or speaking a corresponding wake-up word. In other words, the vehicle-mounted terminal's vehicle-machine interaction system can be woken up by clicking a button or by voice.

[0127] It should be understood that the user who wakes up the vehicle-machine interaction system of the vehicle-mounted terminal and the user who conducts voice interaction with the vehicle-machine interaction system of the vehicle-mounted terminal may be the same user or different users.

[0128] 304. The vehicle-mounted terminal determines target information based on the target voice data and the characteristic data of each of the multiple users.

[0129] 305. The vehicle-mounted terminal pushes target information to the user in the vehicle.

[0130] When the vehicle terminal determines the target information, the target information can be played by voice through the vehicle-machine interaction system, or the target information can be presented on the display screen of the vehicle-machine interaction system for the user to view.

[0131] Using the information recommendation method of the present application, while a vehicle is driving, the vehicle terminal can determine the characteristic data of each user based on the collected information of each of the multiple users in the vehicle. When the user wakes up the vehicle-to-machine interaction system of the vehicle terminal and the vehicle terminal obtains the target voice data of the user in the vehicle, the vehicle terminal will determine the target information that meets the user intention corresponding to the target voice data and the interests of each user in the vehicle based on the target voice data and the characteristic data of each of the multiple users in the vehicle, and push the target information to the user in the vehicle. Compared with the prior art, in which the user in the vehicle needs to engage in multiple rounds of dialogue with the vehicle-to-machine interaction system before obtaining the target information that meets the user's intention and interests, the information recommendation method provided by the embodiment of the present invention can quickly and proactively recommend information to the users in the vehicle. In particular, when the user corresponding to the target voice data is the driver, this information recommendation method can avoid affecting the driver's normal driving and ensure driving safety. Moreover, in the information recommendation process, the embodiment of the present invention takes into account the characteristic data of each user in the vehicle, so that the recommended information can meet the interests of all users in the vehicle, which makes the recommendation results more comprehensive.

[0132] Optionally, the collected information includes capturing images and / or capturing voice data, wherein the captured images include target portrait information of the target user, and the captured voice data includes audio recording data of the target user, where the target user is any one of the multiple users.

[0133] The specific steps of step 302 are introduced below in combination with the specific content of the collected information.

[0134] Combine Figure 3 ,like Figure 4 As shown, when the above-mentioned collected information is a captured image, the above-mentioned step 302 may include the following step 401 or step 402.

[0135] 401. If there is historical portrait information matching the target portrait information in the pre-stored portrait information, the vehicle-mounted terminal determines that the collected information of the target user meets the preset conditions, and obtains the characteristic data of the target user based on the target portrait information and the matching historical portrait information.

[0136] 402. If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user does not meet the preset conditions, and characteristic data of the target user is obtained based on the target portrait information.

[0137] Optionally, after the vehicle-mounted terminal obtains the captured image of the target user, it first uses a pre-stored regional convolutional neural network model to perform image segmentation on the captured image to obtain target portrait information including the target user in the captured image, and the target portrait information includes the target user's facial information. Then, the face recognition model is used to identify the target portrait information to determine whether the target user is a new user or an old user. Finally, if the target user is a new user, a new user portrait is created, and the target portrait information is processed using a pre-stored feature extraction model to obtain the current feature data corresponding to the target user, and the current feature data is stored in the created user portrait; if the target user is an old user, the user portrait corresponding to the target user is found based on the historical portrait information, and the target portrait information is processed using a pre-stored feature extraction model to determine the current feature data corresponding to the target user, and based on the current feature data, the user portrait corresponding to the historical portrait information is updated, and the feature data of the target user is determined based on the updated user portrait.

[0138] It is important to understand that the user profile of an old user includes the user's historical feature data.

[0139] In one embodiment, taking the target user as the user in the main driver's seat as an example, after the vehicle terminal receives the captured image sent by the camera installed on the main driver's seat, the Mask-R-CNN deep neural network model is used to segment the user foreground of the main driver's seat and the boundary background of the main driver's seat. The portrait information of the user in the main driver's seat can be segmented by segmenting the main driver's character and determining the boundary background of the main driver's seat.

[0140] Specifically, the process of segmentation through the Mask-R-CNN deep neural network model includes target detection, target classification, and pixel-level target segmentation. After the captured image is input into the Mask-R-CNN deep neural network model, first, the Mask-R-CNN deep neural network model delineates approximately 2,000 areas to be detected on the captured image; secondly, the Mask-R-CNN deep neural network model extracts features from these 2,000 areas to be detected one by one (in series); thirdly, the support vector machine (SVM) is used to classify the extracted features to obtain the categories of the user on the main driver's seat and the main driver's seat boundary; then, the Mask-R-CNN deep neural network model performs pixel-level target segmentation on the user on the main driver's seat and the main driver's seat boundary, segmenting the user foreground on the main driver's seat and the main driver's seat boundary background. By segmenting the user on the main driver's seat and determining the main driver's seat boundary background, the portrait information of the user in the main driver's seat can be segmented.

[0141] In one embodiment, after isolating the target user's portrait information, the vehicle terminal uses the lightweight SqueezeNet deep neural network model as a facial recognition model to perform facial recognition on the target user. Using the SqueezeNet deep neural network model for facial recognition can reduce the number of parameters by dozens of times, significantly reducing the computing pressure on the vehicle terminal.

[0142] Specifically, the SqueezeNet deep neural network model consists of eight Fire modules, each of which consists of two parts: squeeze and expand. In the SqueezeNet deep neural network model, squeeze represents a squeeze layer, which uses a 1×1 convolution kernel to convolve the feature map of the previous layer, primarily to reduce the feature map's dimensionality. Expand uses the Inception structure, which consists of a 1×1 convolution kernel and a 3×3 convolution kernel, which are then concatenated. After preprocessing the target portrait information, the SqueezeNet deep neural network model extracts the corresponding facial features, performs a facial feature comparison, and outputs the facial recognition result.

[0143] The target user's characteristic data may include the target user's age, gender, language type, and preferences. The language type may include language and accent, and the accent may indicate the user's place of origin.

[0144] In one embodiment, the feature extraction model may include a ResNetV50 deep neural network model and a latent Dirichlet allocation text topic model. The ResNetV50 deep neural network model can be used to extract feature data such as the target user's age, gender, and voice type; the latent Dirichlet allocation text topic model can be used to extract feature data such as the target user's preferences (or interests).

[0145] It is important to understand that the latent Dirichlet distribution text topic model generates a document's text-topic matrix by inputting text and performing a variational-EM algorithm or Gibbs sampling. From this text-topic matrix, the topic features of the text, i.e., the user's interest features, can be extracted.

[0146] Optionally, when the collected information includes only collected voice data, the vehicle-mounted terminal obtaining characteristic data of each user based on the collected information of each user may include: if historical voice data matching the collected voice data exists in the pre-stored voice data, the vehicle-mounted terminal determining that the collected information of the target user meets a preset condition, and obtaining the characteristic data of the target user based on the collected voice data and the matching historical voice data. If historical voice data matching the collected voice data does not exist in the pre-stored voice data, the vehicle-mounted terminal determining that the collected information of the target user does not meet the preset condition, and obtaining the characteristic data of the target user based on the collected voice data.

[0147] Optionally, after the vehicle-mounted terminal obtains the collected voice data of the target user, first, a pre-stored noise reduction model is used to perform noise reduction processing on the collected voice data; secondly, a voice separation model is used to obtain the target user's recorded data included in the collected voice data; then, a voiceprint recognition model is used to perform voiceprint recognition on the target user's recorded data to determine whether the target user is a new user or an old user. Finally, if the target user is a new user, a new user profile is created, and a pre-stored feature extraction model is used to process the target user's recorded data to obtain the current feature data corresponding to the target user, and the current feature data is stored in the created user profile; if the target user is an old user, the user profile corresponding to the target user is found based on the historical voice data, and a pre-stored feature extraction model is used to process the target user's recorded data to determine the current feature data corresponding to the target user, and based on the current feature data, the user profile corresponding to the historical voice data is updated, and the feature data of the target user is determined based on the updated user profile.

[0148] In one embodiment, taking the target user as the user in the main driver's seat as an example, after the vehicle terminal receives the collected voice data transmitted by the microphone installed in the main driver's seat, it first performs noise reduction processing on the collected voice data using the MetricGAN+ deep neural network model, thereby reducing the voices of users in the front passenger and rear seats, as well as other noise in the vehicle. The de-noised voice is then separated using the SepFormer deep neural network model. After the separated voices, the near-field and far-field sound are judged to determine which voice belongs to the user in the main driver's seat.

[0149] There are multiple ways to determine whether the distance between a sound source and the corresponding microphone is far or close. In one possible way, a sound source is considered to be in the far field when the distance from the reference point at the center of the microphone array is much greater than the signal wavelength; otherwise, it is considered to be in the near field. For example, if the distance from the sound source to the center of the array is greater than 2d2 / u (d is the distance between adjacent array elements in a uniform linear array (also known as the array aperture), and u is the wavelength of the highest-frequency speech at the sound source (i.e., the minimum wavelength of the sound source)), then it is a far-field model; otherwise, it is a near-field model. In another possible way, if the sound wave emitted by a microphone is a plane wave, then the sound source corresponding to the sound wave is far away from the microphone, that is, the sound wave is a far-field sound; if the sound wave emitted by a microphone is a spherical wave, then the sound source corresponding to the sound wave is close to the microphone, that is, the sound wave is a near-field sound.

[0150] Specifically, taking the type of sound wave emitted by the microphone as an example, when the sound wave sent by the microphone installed on the main driver's seat to the vehicle terminal is a plane wave, it is judged that the user who emits the sound wave is not the user in the main driver's seat; on the contrary, when the sound wave sent by the microphone installed on the main driver's seat to the vehicle terminal is a spherical wave, it is judged that the user who emits the sound wave is not the user in the main driver's seat.

[0151] It is important to understand that the aforementioned MetricGAN+ deep neural network model uses a neural network to simulate the target evaluation function, and the proxy estimation function is learned from the original score, treating the target evaluation function as a black box. Once the proxy evaluation is trained, it can be used as the loss function of the speech enhancement model. Specifically, the goal of the MetricGAN+ deep neural network model is to optimize the black box metric score. Its training process is similar to that of generative adversarial networks (GANs), using clean speech as the generator and the noise inside the vehicle as the discriminator. The adversarial network is trained to perform speech denoising. The aforementioned SepFormer deep neural network model is a new transformer-based RNN-free speech separation neural network. It learns short-term and long-term dependencies by using the transformer's multi-scale method. It can output separated speech after inputting a mixed speech of multiple people.

[0152] In one embodiment, after separating the target user's recorded data, the vehicle-mounted terminal uses the ResNetV50 deep neural network model as a voiceprint recognition model to perform voiceprint recognition on the target user.

[0153] Specifically, first, the collected voice data is feature extracted through MFCC (Mel-Frequency Cepstral Coefficient) to generate a voice spectrogram; then, the spectrogram is used as input in the ResNetV50 deep neural network model, and 50 conv2d convolution operations are performed on the spectrogram in the ResNetV50 deep neural network model, followed by 4 residual blocks (ResidualBlock), and finally a full connection operation is performed to perform the classification task, the collected voice data is classified into specific people, and the voiceprint recognition results are output.

[0154] It should be understood that when the collected information is voice data, the feature extraction process is the same as the process when the collected information is image capture, and will not be repeated here.

[0155] Optionally, when the collected information includes capturing images and collecting voice data, the vehicle-mounted terminal uses facial recognition as a primary method and voiceprint recognition as a supplement when identifying the target user. When the facial and voiceprint recognition results are greater than a set threshold, i.e., if the target user is determined to be an existing user, the user profile is switched to that user. Conversely, when the facial and voiceprint recognition results are less than a set threshold, i.e., if the target user is determined to be a new user, a user profile for that user is initialized.

[0156] Optionally, the above process of determining the characteristic data of the target user based on the collected information can also be executed in the first server, which can reduce the computing pressure of the vehicle terminal. In this case, there is no feature extraction model in the vehicle terminal.

[0157] Specifically, after the vehicle terminal acquires the collected information of each user in the vehicle, it sends the collected information to the first server. The first server processes each collected information to obtain characteristic data of each user, and sends each user's characteristic data to the vehicle terminal, so that the vehicle terminal acquires the characteristic data of each user. After receiving the characteristic data of all users, the vehicle terminal stores it.

[0158] In one scenario, the vehicle-mounted terminal does not have a relevant model for extracting feature data. In this case, after the vehicle-mounted terminal responds to a wake-up operation but has not yet received the feature data sent by the first server, the feature data of each user is obtained based on the collected information of each user, including: in response to the wake-up operation, if it is determined that the feature data sent by the first server has not been received, if there is historical portrait information matching the target portrait information in the pre-stored portrait information, and there is historical voice data matching the collected voice data in the pre-stored voice data, then the vehicle-mounted terminal determines that the collected information of the target user meets the preset conditions, and obtains the feature data of the target user based on the user portrait corresponding to the historical portrait information and the historical voice data. If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, or there is no historical voice data matching the collected voice data in the pre-stored voice data, then the vehicle-mounted terminal determines that the collected information of the target user does not meet the preset conditions, and uses the preset data as the feature data of the target user.

[0159] That is to say, if the vehicle terminal has obtained the target voice data but has not received the feature data sent by the first server, if the target user is an old user, then the historical feature data of the target user can be directly searched; if the target user is a new user, then the preset data seat can be used as the feature data of the target user.

[0160] Combine Figure 4 ,like Figure 5 As shown, the above step 304 may include the following steps 501 and 502.

[0161] 501. The vehicle-mounted terminal determines the user's intention based on the target voice data.

[0162] After the vehicle-mounted terminal obtains the target voice data, it performs voice processing on the target voice data to determine the user's intention.

[0163] Optionally, after the in-vehicle terminal obtains the target voice data, it first performs voice recognition on the target voice data; then, the target voice data is converted into text information through the DeepSpeech2 deep neural network model; finally, the text information is used to perform intent recognition through a pre-trained language representation model (Bidirectional Encoder Representations from Transformers, BERT) to obtain the user's intent.

[0164] In one embodiment, the speech recognition process may include endpoint detection and MFCC (Mel-Frequency Cepstral Coefficient) feature extraction. Endpoint detection involves detecting the starting and ending points of each speech segment from the middle of a continuous sound, using a dual-threshold front-end detection algorithm. This endpoint detection technique primarily utilizes short-term zero-crossing rate and short-term energy features to determine the starting and ending points of the speech signal. The MFCC feature extraction process includes pre-emphasis, framing, windowing, fast Fourier transform, Mel filtering, and discrete cosine transform.

[0165] In one embodiment, natural language processing is used to identify and output the intent corresponding to the text or keystroke. Intent recognition is based on the BERT model, which uses a BERT model pre-training and fine-tuning architecture to overcome the problem of static word vectors being unable to resolve polysemy. Furthermore, for intent recognition tasks, the BERT model deeply encodes the input sentence to directly obtain the semantic representation of the entire sentence, and the multi-headed attention mechanism makes the semantic information captured by the model more comprehensive.

[0166] 502. The vehicle-mounted terminal determines target information based on the user's intention and characteristic data of each user.

[0167] Combine Figure 5 ,like Figure 6 As shown, the above step 502 may include the following steps 601 to 605.

[0168] 601. The vehicle-mounted terminal determines a target category corresponding to the user's intention.

[0169] After the vehicle terminal determines the user's intention, it uses a pre-stored recommendation model to process the user's intention and determine the target category corresponding to the user's intention.

[0170] For example, the recommendation model can be a latent semantic recommendation model. When it is determined that the user's intention is to listen to music, the corresponding target category is songs. Songs can include multiple attributes, such as fresh music, rock music, and classical music.

[0171] 602. The vehicle-mounted terminal obtains multiple candidate information according to the target category.

[0172] After the vehicle terminal determines the target category, it sends the target category to the second server. After receiving the target category, the second server determines multiple candidate information that meets the target category and sends the multiple candidate information to the vehicle terminal.

[0173] 603. The in-vehicle terminal determines a first coefficient of each candidate information corresponding to the first user based on the feature data of the first user corresponding to the target voice data.

[0174] The first coefficient is used to indicate the first user's interest level in the candidate information.

[0175] Optionally, the process of the on-board terminal determining the first coefficient of each candidate information corresponding to the first user may include the following steps: first, the on-board terminal obtains the weight of each candidate information in each attribute of the target category; then, the on-board terminal determines the recommendation coefficient of each attribute of the target category based on the characteristic data of the first user corresponding to the target voice data, and the recommendation coefficient of each attribute is used to indicate the first user's interest level in the attribute; finally, the on-board terminal determines the first coefficient of each candidate information based on the weight of each candidate information in each attribute and the recommendation coefficient corresponding to each attribute.

[0176] The label of each candidate information can store the weight Q of the candidate information in each attribute of the target category. i The recommendation model pre-stored in the vehicle terminal can analyze the first user's interest level in each attribute corresponding to the target category based on the first user's characteristic data, that is, the recommendation coefficient P i .

[0177] In one embodiment, the following formula (1) may be used to determine the first coefficient corresponding to the first user.

[0178] R i =ΣP i ×Q i (1)

[0179] Among them, P i Indicates the user's interest in each attribute corresponding to the target category, that is, the recommendation coefficient; Q i Represents the weight of each candidate information in each attribute of the target category; R i Indicates the first coefficient of each candidate information corresponding to the first user.

[0180] For example, when the target category is song, each candidate information is a specific song. For a specific song, the weight of the song belonging to the "little fresh" attribute is 0.1, the weight of the song belonging to the "rock" attribute is 0.5, and the weight of the song belonging to the "classical music" attribute is 0.8. The first user's interest in the "little fresh" attribute is 0.7, the interest in the "rock" attribute is 0.6, and the interest in the "classical music" attribute is 0.1. Then the first user's interest in the song, that is, the first coefficient R of the song, is i =0.1×0.7+0.5×0.6+0.8×0.1=0.45.

[0181] The above steps are performed for each song in the song category to determine the first coefficient of each song. Similarly, the above steps can be performed for all categories and will not be repeated here.

[0182] 604. The vehicle-mounted terminal determines a second coefficient of each piece of candidate information corresponding to each second user based on the characteristic data of each second user and a pre-stored recommendation model.

[0183] The second coefficient is used to indicate the interest level of a second user in the candidate information, where the second user is a user other than the first user among the multiple users.

[0184] Optionally, if the collected information of the second user meets the preset conditions, the vehicle-mounted terminal determines the second coefficient of each candidate information corresponding to the second user based on the characteristic data of the second user. If the collected information of the second user does not meet the preset conditions, a preset value is determined as the second coefficient of each candidate information corresponding to the second user.

[0185] In order to consider and balance the interests of all users in the vehicle, all users other than the first user are also considered during the recommendation process. For each second user, if the collected information of the second user meets the preset conditions, that is, the second user is an old user, and the vehicle terminal stores the historical feature data of the second user, then the recommendation model can analyze the second coefficient of each candidate information corresponding to the second user based on the historical feature data of the second user. If the collected information of the second user does not meet the preset conditions, that is, the second user is a new user, then the recommendation model will determine the preset value as the second coefficient of each candidate information corresponding to the second user. The preset value can be an initial value in the recommendation model.

[0186] In one embodiment, the following formula (2) may be used to determine a second coefficient corresponding to a second user.

[0187] W i =∑b i ×w i (2)

[0188] Among them, b i represents the initial interest level of a second user in each candidate information; w i represents the preference weight of a second user for each candidate information; W i A second coefficient representing each piece of candidate information corresponding to a second user.

[0189] For example, a second user's initial interest level in a song is 0.8, and the preference weight is 0.7, then the second user's final interest level in the song, that is, the second coefficient, is 0.56.

[0190] 605. The vehicle-mounted terminal determines target information from a plurality of candidate information according to the determined first coefficient and second coefficient.

[0191] Optionally, the process of the on-board terminal determining the target information based on the first coefficient and all the second coefficients may include the following steps: first, the on-board terminal determines the target coefficient of each candidate information, i.e., inside the vehicle, based on the first coefficient and all the second coefficients; then, the on-board terminal sorts the multiple candidate information according to all the target coefficients, and generates the target information based on the sorting results of the multiple candidate information.

[0192] For example, the target coefficient may be determined using the following formula (3).

[0193]

[0194] Among them, W ui represents the second coefficient of the u-th user, where u ranges from [1, m], m is a positive integer, and the value of m is related to the number of users in the vehicle; S i represents the target coefficient of each candidate information for u users in the vehicle.

[0195] For example, if there is only one first user and one second user in the vehicle, the recommendation model ultimately determines the target coefficient for the specific song as 1.01. After the recommendation model determines the recommendation coefficient for each candidate information, it sorts all the target coefficients to obtain the sorted result of all the candidate information. The sorted candidate information is the target information.

[0196] When the second user is a new user, the preset value may be determined as the above b i and w i When the second user is an old user, the above b can be determined based on the historical feature data of the second user. i and w i .

[0197] In one embodiment, during the training of the recommendation model, the above b can be determined by minimizing the loss function. i and w i When the second user is an old user, the following formula (4) can be used to determine the above b i and w i .

[0198]

[0199] Among them, C' represents the loss value; S ui represents the real sample data when training the recommendation model, S u ' iIndicates the target coefficient's estimated value; It is a regularization term used to prevent overfitting, and λ is a regularization parameter.

[0200] Optionally, after the first user selects a candidate from the target information, the in-vehicle terminal uploads the selection result as feedback information to the first server. The first server also pre-stores a recommendation model. After receiving the feedback information, the first server trains the recommendation model based on the feedback information to achieve an updated recommendation model iteration, and then sends the updated recommendation model to the in-vehicle terminal. In this way, the in-vehicle terminal can recommend more accurate target information to the user in the vehicle during the next recommendation process.

[0201] The above mainly introduces the solution provided by the embodiment of the present invention from the perspective of the device. It can be understood that in order to realize the above functions, the device includes a hardware structure and / or software module corresponding to the execution of each function. It should be easily appreciated by those skilled in the art that, in combination with the algorithm steps of each example described in the embodiments disclosed herein, the present invention can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0202] Figure 7 A possible schematic diagram of the composition of the information recommendation device 700 involved in the above embodiment is shown. Figure 7 As shown, the information recommendation device 700 may include: an acquisition unit 701 , a determination unit 702 and a sending unit 703 .

[0203] The acquisition unit 701 is configured to acquire collected information from each of the multiple users in the vehicle. The determination unit 702 is configured to determine characteristic data of each user based on the collected information of each user. The acquisition unit 701 is further configured to acquire target voice data in response to a wake-up operation. The determination unit 702 is further configured to determine target information based on the target voice data and the characteristic data of each of the multiple users. The sending unit 703 is configured to push the target information to the users in the vehicle.

[0204] Optionally, the determining unit 702 is specifically configured to:

[0205] The user intention is determined based on the target voice data, and the target information is determined based on the user intention and the characteristic data of each user.

[0206] Optionally, the determining unit 702 is specifically configured to:

[0207] Determine a target category corresponding to the user intention, determine multiple candidate information based on the target category and the user intention, and determine a first coefficient of each candidate information corresponding to the first user based on the feature data of the first user corresponding to the target voice data, the first coefficient being used to indicate the first user's interest level in the candidate information, and determine a second coefficient of each candidate information corresponding to each second user based on the feature data of each second user and a pre-stored recommendation model, the second coefficient being used to indicate the second user's interest level in the candidate information, the second user being a user other than the first user among the multiple users, and determine the target information from the multiple candidate information based on the determined first coefficient and second coefficient.

[0208] Optionally, the determining unit 702 is specifically configured to:

[0209] If the collected information of the second user meets the preset conditions, the second coefficient of each candidate information corresponding to the second user is determined based on the characteristic data of the second user; if the collected information of the second user does not meet the preset conditions, the preset value is determined as the second coefficient of each candidate information corresponding to the second user.

[0210] Optionally, the determining unit 702 is specifically configured to:

[0211] Obtain the weight of each candidate information in each attribute of the target category, determine the recommendation coefficient of each attribute of the target category based on the feature data of the first user corresponding to the target voice data, the recommendation coefficient of each attribute is used to indicate the degree of interest of the first user in the attribute, and determine the first coefficient of each candidate information based on the weight of each candidate information in each attribute and the recommendation coefficient corresponding to each attribute.

[0212] Optionally, the collected information includes a captured image, the captured image includes target portrait information of a target user, and the target user is any user among a plurality of users. The determination unit 702 is specifically configured to:

[0213] If there is historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user meets the preset conditions, and the characteristic data of the target user is determined based on the target portrait information and the matching historical portrait information; if there is no historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user does not meet the preset conditions, and the characteristic data of the target user is obtained based on the target portrait information.

[0214] Optionally, the determining unit 702 is specifically configured to:

[0215] Determine current feature data based on target portrait information; update user portraits corresponding to historical portrait information based on current feature data; and determine feature data of the target user based on the updated user portraits.

[0216] Optionally, the collected information includes collected voice data, and the collected voice data includes recorded data of the target user. The determination unit 702 is specifically configured to:

[0217] If there is historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user meets the preset conditions, and the characteristic data of the target user is determined based on the collected voice data and the matching historical voice data; if there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user meets the non-preset conditions, and the characteristic data of the target user is obtained based on the collected voice data.

[0218] Optionally, the sending unit 703 is further configured to send the collected information acquired by the acquiring unit 701 to the server, where the collected information is used by the server to determine the feature data.

[0219] Optionally, the determining unit 702 is specifically configured to:

[0220] In response to the wake-up operation, if it is determined that no feature data sent by the server has been received, if there is historical portrait information matching the target portrait information in the pre-stored portrait information, and there is historical voice data matching the collected voice data in the pre-stored voice data, then determining that the collected information of the target user meets the preset conditions, and obtaining the feature data of the target user based on the user portrait corresponding to the historical portrait information and the historical voice data;

[0221] If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, or there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user does not meet the preset conditions, and the preset data is used as the feature data of the target user.

[0222] Optionally, the acquisition unit 701 is specifically configured to:

[0223] The collected information of the user of each seat is obtained through the collection device corresponding to each seat in the vehicle, and the collected information of each user among the multiple users is obtained.

[0224] Of course, the information recommendation device 700 provided by the embodiment of the present invention includes but is not limited to the above modules.

[0225] In actual implementation, the acquisition unit 701, the determination unit 702 and the sending unit 703 can be composed of Figure 2The processor 21 shown calls the program code in the memory 22 to implement the process. Figures 3 to 6 The description of the information recommendation method shown is not repeated here.

[0226] Another embodiment of the present invention further provides a computer-readable storage medium having computer instructions stored therein. When the computer instructions are executed on the information recommendation device 700, the information recommendation device 700 executes each step executed by the vehicle in the method flow shown in the above method embodiment.

[0227] Another embodiment of the present invention provides a chip system, which is applied to an information recommendation device 700. The chip system includes one or more interface circuits and one or more processors 21. The interface circuits and processors 21 are interconnected via circuits. The interface circuits are configured to receive signals from the memory 22 of the information recommendation device 700 and send these signals to the processors 21. The signals include computer instructions stored in the memory 22. When the processors 21 execute the computer instructions, the information recommendation device 700 performs the steps described in the method flow of the above method embodiment.

[0228] In another embodiment of the present invention, a computer program product is provided. The computer program product includes instructions. When the instructions are executed on the information recommendation device 700, the information recommendation device 700 executes each step performed by the information recommendation device 700 in the method flow shown in the above method embodiment.

[0229] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using a software program, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer-executable instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more media integrated therein. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a DVD), or a semiconductor medium (eg, a solid state disk (SSD)).

[0230] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions within the technical scope disclosed by the present invention shall be covered by the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. An information recommendation method, characterized in that: include: Obtaining collected information of each of a plurality of users in the vehicle; Obtain characteristic data of each user based on the collected information of each user; In response to the wake-up operation, acquiring target voice data; determining user intent based on the target voice data; Determining a target category corresponding to the user intention; obtaining multiple candidate information based on the target category; determining, based on feature data of a first user corresponding to the target voice data, a first coefficient for each piece of candidate information corresponding to the first user, the first coefficient being used to indicate a degree of interest of the first user in the candidate information; Determining, based on the characteristic data of each second user and a pre-stored recommendation model, a second coefficient for each piece of candidate information corresponding to each second user, including: if the collected information of the second user meets a preset condition, determining, based on the characteristic data of the second user, the second coefficient for each piece of candidate information corresponding to the second user; if the collected information of the second user does not meet the preset condition, determining a preset value as the second coefficient for each piece of candidate information corresponding to the second user, the second coefficient being used to indicate the second user's level of interest in the candidate information, the second user being a user among the multiple users excluding the first user; determining target information from the plurality of candidate information according to the determined first coefficient and the second coefficient; The target information is pushed to a user in the vehicle.

2. The information recommendation method according to claim 1, characterized in that: The determining, based on the feature data of the first user corresponding to the target voice data, a first coefficient of each candidate information corresponding to the first user includes: Obtain the weight of each candidate information in each attribute of the target category; determining, based on the feature data of the first user corresponding to the target voice data, a recommendation coefficient for each attribute of the target category, wherein the recommendation coefficient for each attribute is used to indicate the first user's interest in the attribute; The first coefficient of each candidate information is determined according to the weight of each candidate information in each attribute and the recommendation coefficient corresponding to each attribute.

3. The information recommendation method according to claim 1 or 2, characterized in that: The collected information includes a captured image, the captured image includes target portrait information of a target user, and the target user is any user among the multiple users; The step of obtaining characteristic data of each user based on the collected information of each user includes: If there is historical portrait information matching the target portrait information in the pre-stored portrait information, determining that the collected information of the target user meets the preset conditions, and acquiring the characteristic data of the target user based on the target portrait information and the matching historical portrait information; If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, it is determined that the collected information of the target user does not meet the preset condition, and characteristic data of the target user is obtained based on the target portrait information.

4. The information recommendation method according to claim 3, characterized in that: The acquiring of characteristic data of the target user according to the target portrait information and the matched historical portrait information includes: Determining current feature data based on the target portrait information; Update the user portrait corresponding to the historical portrait information according to the current feature data; Determine the characteristic data of the target user based on the updated user portrait.

5. The information recommendation method according to claim 1 or 2, characterized in that: The collected information includes collected voice data, and the collected voice data includes recording data of a target user, and the target user is any user among the multiple users; The step of obtaining characteristic data of each user based on the collected information of each user includes: If there is historical voice data matching the collected voice data in the pre-stored voice data, determining that the collected information of the target user meets the preset conditions, and obtaining feature data of the target user based on the collected voice data and the matching historical voice data; If there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user does not meet the preset condition, and characteristic data of the target user is obtained based on the collected voice data.

6. The information recommendation method according to claim 1 or 2, characterized in that: The collected information includes captured images and collected voice data, the captured images include target portrait information of a target user, and the collected voice data includes recorded data of a target user, where the target user is any one of the multiple users; The information recommendation method further includes: sending collected information to a server, wherein the collected information is used by the server to determine feature data; The step of obtaining characteristic data of each user based on the collected information of each user includes: In response to the wake-up operation, if it is determined that no feature data sent by the server has been received, if historical portrait information matching the target portrait information exists in the pre-stored portrait information, and historical voice data matching the collected voice data exists in the pre-stored voice data, determining that the collected information of the target user meets a preset condition, and acquiring the feature data of the target user based on the user portrait corresponding to the historical portrait information and the historical voice data; If there is no historical portrait information matching the target portrait information in the pre-stored portrait information, or there is no historical voice data matching the collected voice data in the pre-stored voice data, it is determined that the collected information of the target user does not meet the preset conditions, and the preset data is used as the feature data of the target user.

7. An information recommendation device, characterized in that: include: an acquiring unit, configured to acquire collected information of each of a plurality of users in the vehicle; a determination unit, configured to determine characteristic data of each user based on the collected information of each user; The acquisition unit is further configured to acquire target voice data in response to a wake-up operation; The determining unit is further configured to determine a user intention based on the target voice data; determine a target category corresponding to the user intention; obtain a plurality of candidate information based on the target category; and determine a first coefficient for each candidate information corresponding to the first user based on feature data of the first user corresponding to the target voice data, the first coefficient being used to indicate a degree of interest of the first user in the candidate information; determining, based on the characteristic data of each second user and a pre-stored recommendation model, a second coefficient for each candidate information corresponding to each second user, including: if the collected information of the second user meets a preset condition, determining, based on the characteristic data of the second user, the second coefficient for each candidate information corresponding to the second user; if the collected information of the second user does not meet the preset condition, determining a preset value as the second coefficient for each candidate information corresponding to the second user, the second coefficient being used to indicate the second user's level of interest in the candidate information, the second user being a user among the multiple users excluding the first user; and determining target information from the multiple candidate information based on the determined first coefficient and the second coefficient; A sending unit is used to push the target information to the user in the vehicle.

8. A vehicle, characterized in that: The vehicle includes: a processor and a memory; the memory is used to store computer program code, and the computer program code includes computer instructions; when the processor executes the computer instructions, the vehicle executes the information recommendation method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The method comprises computer instructions, which, when executed on a vehicle, cause the vehicle to execute the information recommendation method according to any one of claims 1 to 6.

Images