In-vehicle voice interaction method and device, storage medium and in-vehicle infotainment system

Abstract

The invention relates to the technical field of man-machine conversation, and discloses an in-vehicle voice interaction method and device, a storage medium and an in-vehicle infotainment system, and the method comprises the steps: carrying out the semantic analysis of a current voice instruction through a first voice model when the current voice instruction is obtained, and enabling the current voice instruction to be a voice instruction which cannot be recognized by a second voice model; determining a target intention corresponding to the current voice instruction according to the analysis result, and determining a corresponding third voice model according to the target intention; broadcasting a preset wake-up word corresponding to the third voice model, wherein the preset wake-up word is used for waking up the third voice model; and generating a corresponding voice submission instruction based on the target intention, and broadcasting the voice submission instruction to enable a third voice model to execute the voice submission instruction. Compared with the prior art, the target voice model can be automatically awakened according to the current voice instruction, and the instruction is broadcasted, so that the user operation is simplified, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the field of human-computer dialogue technology, and in particular to an in-vehicle voice interaction method, device, storage medium and vehicle-computer system. Background Art

[0002] Currently, when a user is conversing with the voice model of an in-vehicle application (Application, APP), if the user needs to switch to the voice model of another APP for conversation, he needs to wait for the current voice model to exit and then wake up the target voice model; for example, if the user is conversing with the voice model of a navigation APP but wants to switch to the voice model of a music APP for conversation, the user needs to wait for the voice model of the navigation APP to exit and then wake up the voice model of the music APP, which makes the user operation more cumbersome.

[0003] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0004] The main purpose of the present invention is to provide an in-vehicle voice interaction method, device, storage medium and vehicle-machine system, aiming to solve the technical problem in the prior art that when switching from one voice model to another, the user needs to wait for the current voice model to exit and then wake up the target voice model, which leads to cumbersome user operations.

[0005] To achieve the above object, the present invention provides an in-vehicle voice interaction method, the method is applied to a vehicle system, the vehicle system has at least a first voice model, a second voice model, and a third voice model, the method comprises the following steps:

[0006] When a current voice instruction is acquired, semantic analysis is performed on the current voice instruction by using the first voice model, and the current voice instruction is a voice instruction that cannot be recognized by the second voice model;

[0007] Determine the target intention corresponding to the current voice instruction according to the analysis result, and determine the corresponding third voice model according to the target intention;

[0008] Broadcasting a preset wake-up word corresponding to the third voice model, where the preset wake-up word is used to wake up the third voice model;

[0009] A corresponding voice reporting instruction is generated based on the target intention, and the voice reporting instruction is broadcast so that the third voice model executes the voice reporting instruction.

[0010] Optionally, before the step of performing semantic analysis on the current voice instruction by using the first voice model when the current voice instruction is acquired, the method further includes:

[0011] Acquire a function label corresponding to each voice model, and determine a support intent corresponding to each voice model based on each function label;

[0012] Constructing a preset mapping relationship according to the support intentions corresponding to each of the voice models;

[0013] The step of determining the corresponding third speech model according to the target intention comprises:

[0014] The third speech model corresponding to the target intent is determined based on the preset mapping relationship.

[0015] Optionally, before the step of obtaining the function label corresponding to each speech model, the method further includes:

[0016] When the voice model to be bound successfully executes the received voice instruction, obtaining the execution content corresponding to the voice model to be bound;

[0017] Generate a corresponding function tag based on the execution content, and bind the function tag to the voice model to be bound.

[0018] Optionally, the step of determining the third speech model corresponding to the target intent based on the preset mapping relationship includes:

[0019] Determine a list of optional voice models corresponding to the target intent based on the preset mapping relationship;

[0020] Performing intention improvement on the target intention according to the analysis result;

[0021] The third voice model is determined from the list of optional voice models through the refined target intent.

[0022] Optionally, the step of determining the target intention corresponding to the current voice instruction according to the analysis result includes:

[0023] Determine the primary intention and the secondary intention corresponding to the current voice command according to the analysis result, and use the primary intention as the target intention;

[0024] The step of perfecting the target intention according to the analysis result comprises:

[0025] The target intent is refined according to the secondary intent.

[0026] Optionally, before the step of generating a corresponding voice reporting instruction based on the target intention, the method further includes:

[0027] Determining whether a latest voice command is received within a preset time period;

[0028] If so, a latest reporting instruction is generated according to the latest voice instruction, and the latest reporting instruction is broadcasted so that the third voice model executes the latest reporting instruction.

[0029] Optionally, before the step of generating the latest reporting instruction according to the latest voice instruction, the method further includes:

[0030] Determining whether the latest voice instruction matches the third voice model;

[0031] If yes, the step of generating the latest reporting instruction according to the latest voice instruction is executed.

[0032] In addition, to achieve the above-mentioned purpose, the present invention also proposes an in-vehicle voice interaction device, which includes: a memory, a processor, and an in-vehicle voice interaction program stored in the memory and executable on the processor, and the in-vehicle voice interaction program is configured to implement the steps of the in-vehicle voice interaction method described above.

[0033] In addition, to achieve the above-mentioned purpose, the present invention also proposes a storage medium, on which an in-vehicle voice interaction program is stored. When the in-vehicle voice interaction program is executed by a processor, the steps of the in-vehicle voice interaction method described above are implemented.

[0034] In addition, to achieve the above-mentioned purpose, the present invention further provides a vehicle system, the vehicle system comprising:

[0035] A semantic analysis module, configured to perform semantic analysis on a current voice instruction by using a first voice model when a current voice instruction is acquired, wherein the current voice instruction is a voice instruction that cannot be recognized by the second voice model;

[0036] A model determination module, used to determine the target intention corresponding to the current voice instruction according to the analysis result, and determine the corresponding third voice model according to the target intention;

[0037] A wake-up word broadcasting module, used to broadcast a preset wake-up word corresponding to the third voice model, and the preset wake-up word is used to wake up the third voice model;

[0038] An instruction generation module is used to generate a corresponding voice reporting instruction based on the target intention, and broadcast the voice reporting instruction so that the third voice model executes the voice reporting instruction.

[0039] The present invention provides a method, device, storage medium and vehicle-mounted system for in-vehicle voice interaction. The method performs semantic analysis on the current voice instruction through the first voice model when the current voice instruction is obtained, and the current voice instruction is a voice instruction that cannot be recognized by the second voice model; determines the target intention corresponding to the current voice instruction according to the analysis result, and determines the corresponding third voice model according to the target intention; broadcasts the preset wake-up word corresponding to the third voice model, and the preset wake-up word is used to wake up the third voice model; generates a corresponding voice reporting instruction based on the target intention, and broadcasts the voice reporting instruction so that the third voice model executes the voice reporting instruction. Because the present invention can determine the corresponding third voice model according to the current voice instruction after obtaining the current voice instruction, and wake up the third voice model by playing the corresponding preset wake-up word according to the third voice model, and finally broadcasts the corresponding voice reporting instruction so that the third voice model performs voice interaction according to the voice reporting instruction. Compared with the existing method that requires the user to wait for the current voice model to exit before waking up the target voice model, the present invention can automatically wake up the target voice model according to the current voice instruction and broadcast the instruction without user operation, thereby simplifying the user operation and improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 A schematic diagram of the structure of an in-vehicle voice interaction device in a hardware operating environment involved in an embodiment of the present invention;

[0041] Figure 2 It is a flowchart of the first embodiment of the in-vehicle voice interaction method of the present invention;

[0042] Figure 3 It is a flowchart of a second embodiment of the in-vehicle voice interaction method of the present invention;

[0043] Figure 4 A schematic diagram of a flow chart of a third embodiment of the in-vehicle voice interaction method of the present invention;

[0044] Figure 5 This is a structural block diagram of the first embodiment of the vehicle system of the present invention.

[0045] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0046] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0047] Reference Figure 1 , Figure 1 The figure is a schematic diagram of the structure of the in-vehicle voice interaction device in the hardware operating environment involved in the embodiment of the present invention.

[0048] like Figure 1 As shown, the in-vehicle voice interaction device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a wireless fidelity (Wireless-Fidelity, Wi-Fi) interface). The memory 1005 may be a high-speed random access memory (Random Access Memory, RAM), or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk storage. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0049] Those skilled in the art will understand that Figure 1 The structure shown in does not constitute a limitation on the in-vehicle voice interaction device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0050] like Figure 1 As shown, the memory 1005 as a storage medium may include an operating system, a network communication module, a user interface module, and an in-vehicle voice interaction program.

[0051] exist Figure 1 In the in-vehicle voice interaction device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the in-vehicle voice interaction device of the present invention can be set in the in-vehicle voice interaction device, and the in-vehicle voice interaction device calls the in-vehicle voice interaction program stored in the memory 1005 through the processor 1001, and executes the in-vehicle voice interaction method provided by the embodiment of the present invention.

[0052] It should be noted that, currently, when a user is conversing with the voice model of an in-vehicle application (Application, APP), if the user needs to switch to the voice model of another APP for conversation, he needs to wait for the current voice model to exit and then wake up the target voice model; for example, if the user is conversing with the voice model of the navigation APP but wants to switch to the voice model of the music APP for conversation, the user needs to wait for the voice model of the navigation APP to exit and then wake up the voice model of the music APP, which makes the user operation more cumbersome.

[0053] Therefore, in order to solve the above defects, after receiving the current voice command, this embodiment can determine the corresponding third voice model according to the current voice command, and play the corresponding preset wake-up word according to the third voice model to wake up the third voice model, and finally broadcast the corresponding voice reporting instruction, so that the third voice model can perform voice interaction according to the voice reporting instruction. Compared with the existing system that requires the user to wait for the current voice model to exit before waking up the target voice model, this embodiment can automatically wake up the target voice model according to the current voice command and broadcast the instruction without user operation, thereby simplifying user operation and improving user experience.

[0054] For ease of understanding, refer to Figures 2 to 5 The embodiments of the present invention and the following embodiments are described in detail.

[0055] Reference Figure 2 , Figure 2 Schematic diagram of the flow of the first embodiment of the in-vehicle voice interaction method of the present invention.

[0056] In this embodiment, the in-vehicle voice interaction method is applied to a vehicle system, and the vehicle system has at least a first voice model, a second voice model, and a third voice model. The method includes the following steps:

[0057] Step S10: When a current voice instruction is acquired, semantic analysis is performed on the current voice instruction through the first voice model, and the current voice instruction is a voice instruction that cannot be recognized by the second voice model.

[0058] It is understandable that the method of this embodiment can be applied in the scenario of human-computer dialogue with the in-vehicle APP. The execution subject of the method of this embodiment can be a computing service device with in-vehicle voice interaction, network communication and program running functions, such as a car system, etc., and can also be other electronic devices that implement the same or similar functions. The following is an explanation of this embodiment and the following embodiments using the above-mentioned in-vehicle voice interaction device (referred to as the device).

[0059] It should be understood that the above-mentioned first voice model can be a voice model at the system level in the vehicle system, and the above-mentioned second voice model and the third voice model can be voice models corresponding to different APPs in the vehicle system, wherein the second voice model can be a model that is currently running and cannot recognize the above-mentioned current voice command, and the third voice model can be a model that can recognize the current voice command. At least the above-mentioned first voice model at the system level, as well as the second voice model and the third voice model corresponding to two different APPs can be set in the above-mentioned vehicle system. Of course, voice models corresponding to more APPs can also be set, and this embodiment does not limit this.

[0060] In this embodiment, a preset voice model can be set in the above-mentioned device (i.e., the above-mentioned vehicle system) as the above-mentioned first voice model. Under normal circumstances, the above-mentioned preset voice model does not run, but if the current voice model that communicates with the user cannot process the current voice command, the above-mentioned voice model will be awakened.

[0061] Exemplarily, if the above-mentioned preset voice model (i.e., the above-mentioned first voice model) is recorded as model A, and the user is currently conversing with the voice model B model (i.e., the above-mentioned second voice model) of the navigation APP, since the general voice model will stay for a period of time after processing a voice command issued by the user when conversing with the user, so that the user can issue another voice command, for example, after the user issues a voice command "navigate to sports center a" to model B, model B may stay for a period of time after processing the above voice command. If the user issues "play song c by singer b" at this time, and since the conduction APP corresponding to model B does not have the song playing function, and model B cannot recognize the voice command, the above-mentioned device can use the voice command "play song c by singer b" as the above-mentioned current voice command.

[0062] It should also be noted that the above-mentioned device can monitor the recognition results of voice commands recognized by the voice model in each APP in real time, which can be achieved by collecting the reply results of the second voice model. For example, after receiving "play song c by singer b", model B cannot recognize it, and then model B can broadcast "sorry, this application does not support this function yet" and other similar execution failure words. When the above-mentioned device monitors the above-mentioned execution failure words, it can be determined that the second voice model cannot recognize the current voice command, and then the above-mentioned preset voice model can be triggered, that is, triggering model A.

[0063] After the preset voice model is triggered, the current voice command can be input into it, and then the preset voice model can perform semantic analysis on the current voice command to analyze the specific meaning of the current voice command.

[0064] In a specific implementation, the above-mentioned device can monitor the currently running second voice model and obtain the processing result of the second voice model processing the current voice command. If the processing result is an execution failure, the preset voice large model can be awakened, and the current voice command can be input into the preset voice large model, and the current voice command can be semantically analyzed through the preset voice large model.

[0065] Step S20: Determine the target intention corresponding to the current voice instruction according to the analysis result, and determine the corresponding third voice model according to the target intention.

[0066] It is understandable that after analyzing the current voice command through the above-mentioned preset large voice model, the corresponding target intent can be obtained, and the above-mentioned device can determine the third voice model that matches the target intent according to the target intent to achieve the target intent. Exemplarily, after semantic analysis of the above-mentioned voice instruction of "play song c by singer b", the target intent that can be obtained is "play music, and play song c by singer b", and then the above-mentioned device can search for the APP for playing music from all voice models, and determine which voice model corresponds to the APP for playing music as the above-mentioned third voice model. For the convenience of subsequent explanation, this embodiment will record the voice model of the APP that plays music as C model, and C model is the above-mentioned third voice model.

[0067] In a specific implementation, the above-mentioned device can determine the target intention corresponding to the current voice instruction according to the analysis result, and determine the third voice model corresponding to the matching target application based on the target intention.

[0068] Step S30: broadcasting a preset wake-up word corresponding to the third voice model, wherein the preset wake-up word is used to wake up the third voice model.

[0069] Since different voice models correspond to different wake-up words, the above-mentioned device may store the wake-up words corresponding to each voice model. In a specific implementation, after the above-mentioned device determines the third voice model, it may determine the preset wake-up word corresponding to the third voice model and broadcast it. For example, if the preset wake-up word corresponding to the above-mentioned C model is "Hello, Little C", the above-mentioned device may broadcast the words "Hello, Little C" through the speaker in the car to wake up the C model.

[0070] Step S40: Generate a corresponding voice reporting instruction based on the target intention, and broadcast the voice reporting instruction, so that the third voice model performs voice interaction according to the voice reporting instruction.

[0071] The above-mentioned voice reporting instruction can be a voice instruction broadcasted by the above-mentioned device on behalf of the user. In this embodiment, when the C model is awakened, the above-mentioned device can generate a corresponding voice reporting instruction according to the target intention, that is, generate a corresponding voice reporting instruction according to "play music, and play song c by singer b" and broadcast it.

[0072] Among them, when generating the corresponding voice reporting instruction according to the target intent, the above device can directly repeat the current voice instruction once, that is, repeatedly broadcast "play song c by singer b" to model C once. However, it should be emphasized that considering that there may be some interfering instructions in the current voice instruction, a command generation template can also be set in the above device, and the corresponding voice reporting instruction can be generated according to the target intent through the command generation template.

[0073] Exemplarily, the instruction generation template may be in the form of a verb plus a noun, that is, the device may identify the target intent, extract the verb (i.e., play) and noun (i.e., music, song c by singer b), taking into account the possibility of multiple verbs or nouns, that is, the nouns may include "music" and "song c by singer b", and then the device may integrate the multiple verbs or nouns that appear to generate the most detailed verbs or nouns. For example, after the integration of the nouns, "song c by singer b" can be generated.

[0074] The integration method may be to select the noun with the largest number of characters, or to integrate the nouns according to their semantics. The specific method is not limited in this embodiment.

[0075] After integration, the above device can obtain the verb "play" and the noun "song c by singer b", and then substitute the above instruction generation template to obtain the voice reporting instruction of "play song c by singer b" and broadcast it. After receiving the voice reporting instruction of "play song c by singer b", model C can perform the corresponding operation. Therefore, the whole process does not require user switching, which simplifies user operation and improves user experience.

[0076] Further, considering that after the above-mentioned device wakes up the corresponding third voice model, the user may not need to recognize the previous current voice command and issues a new voice command. For example, the user re-issues the voice command "play song e by singer d". In order to respond to the latest voice command issued by the user in a timely manner, in this embodiment, before the step of generating the corresponding voice report command based on the target intent, it also includes:

[0077] Step S401: Determine whether a latest voice command is received within a preset time period.

[0078] Step S402: If yes, then generate a latest reporting instruction according to the latest voice instruction, and broadcast the latest reporting instruction, so that the third voice model executes the latest reporting instruction.

[0079] After waking up the third voice model, the above-mentioned device will stay for a period of time and continue to listen to whether the user issues a new voice command during the preset time period. If so, the previous current voice command can be replaced with the latest voice command, and the latest reporting command can be generated through the command generation template according to the latest voice command for broadcast, so that the third voice model can execute the latest voice command.

[0080] If the latest voice command is not received within the preset time period, the above device will continue to broadcast the voice reporting command generated by the current voice command.

[0081] It should be emphasized that before replacing the previous current voice instruction with the latest voice instruction, the above device may also determine whether the latest voice instruction is executed by the third voice model, specifically:

[0082] Before the above step of generating the latest proxy reporting instruction according to the latest voice instruction, it also includes: determining whether the latest voice instruction matches the third voice model; if so, executing the step of generating the latest proxy reporting instruction according to the latest voice instruction.

[0083] It is understandable that the above judgment method can still be used to determine the target intention corresponding to the latest semantic instruction. The specific judgment process is similar to the above, and this embodiment will not elaborate on it.

[0084] For example, if the latest voice command is "play song e by singer d", it can be determined that the latest voice command matches the C model and is still executed by the C model, indicating that the user's intention has changed. Then the above device performs a voice command replacement operation to replace "play song c by singer b" with "play song e by singer d", and generates the latest reporting command according to the voice command "play song e by singer d";

[0085] But for example, if the latest voice command is "adjust the air conditioner to 18 degrees", since the C model is the voice model corresponding to playing music, the voice command cannot be executed, and the above-mentioned device can determine that it is not executed by the C model, which means that the user's intention is to add, not to change. Therefore, the above-mentioned device can first generate a voice reporting instruction according to the voice command of "play song e by singer d" and broadcast it to enable the C model to execute it, and continue to broadcast the voice command of "adjust the air conditioner to 18 degrees" after the C model is executed. Since the C model cannot recognize it, it can return to the above-mentioned step S10 and still realize automatic switching of the voice model.

[0086] In a specific implementation, the above-mentioned device can determine whether the user's target intention has changed or a new target intention has been added based on the latest voice command received, and generate the latest reporting instruction based on the latest voice command when the target intention has changed. When the target intention has been added, it still generates a voice reporting instruction based on the current voice command. After broadcasting the voice reporting instruction, it continues to broadcast the latest reporting instruction to achieve automatic switching of voice models.

[0087] The above-mentioned device of this embodiment can monitor the second voice model currently running, obtain the processing result of the second voice model processing the current voice instruction, and if the processing result is execution failure, the preset voice model can be awakened, and the current voice instruction can be input into the preset voice model, and the current voice instruction can be semantically analyzed by the preset voice model; the target intent corresponding to the current voice instruction is determined according to the analysis result, and the third voice model corresponding to the matching target application is determined based on the target intent; the preset wake-up word corresponding to the third voice model is determined and broadcasted; after the third voice model is awakened, the voice reporting instruction generated according to the target intent is broadcasted so that the third voice model executes the current voice instruction. Compared with the existing method that requires the user to wait for the current voice model to exit before waking up the target voice model, this embodiment can automatically wake up the target voice model according to the current voice instruction and broadcast the instruction without user operation, thereby simplifying the user operation and improving the user experience.

[0088] refer to Figure 3 , Figure 3 Schematic diagram of the flow chart of the second embodiment of the in-vehicle voice interaction method of the present invention.

[0089] In order to accurately select the corresponding target semantic model according to the target intention, such as Figure 3 As shown, in this embodiment, before the step of performing semantic analysis on the current voice instruction when receiving the current voice instruction, the step further includes:

[0090] Step S01: Acquire the function label corresponding to each speech model, and determine the support intention corresponding to each speech model based on each function label.

[0091] It should be noted that the above function labels can be labels corresponding to the functions of the voice model. Based on the above example, since model B is a voice model in the communication APP, the function labels corresponding to model B may include but are not limited to "navigation, public transportation, and scenic spot search", and model C is a voice model in the music APP, and the function labels corresponding to model C may include but are not limited to "play songs, play music, play radio stations", etc. The function labels corresponding to specific voice models can be set according to the functions corresponding to the voice models, and this embodiment does not limit this.

[0092] It is understandable that the above-mentioned supported intentions can be user intentions that can be supported by each function label. For example, the function label corresponding to model C is "play songs, play music, play radio", then the user intentions that can be supported by model C are "play songs, play music, play radio".

[0093] Furthermore, in order to determine the function labels of each speech model, in this embodiment, before the step of obtaining the function labels corresponding to each speech model, the following steps are further included:

[0094] Step S001: When the voice model to be bound successfully executes the received voice instruction, the execution content corresponding to the voice model to be bound is obtained.

[0095] The above-mentioned device can monitor the running results of a certain voice model (i.e., the above-mentioned voice model to be bound) when the voice model is running. For example, when a user is having a conversation with the C model, if the content of the conversation is that the user issues a voice command "play song e by singer d", the C model can successfully execute the voice command. After determining that the C model has successfully executed the voice command, the above-mentioned device can obtain what content the C model has specifically executed (i.e., the above-mentioned "play song e by singer d") as the above-mentioned execution content.

[0096] Step S002: Generate a corresponding function tag based on the execution content, and bind the function tag to the voice model to be bound.

[0097] The above-mentioned device can also analyze the execution content, determine the functions involved therein, generate a corresponding function tag, and bind the function tag to the currently running voice model. Based on the above example, in this embodiment, the above-mentioned device can analyze "play song e by singer d", determine that the verb is "play" and the noun is "song e by singer d", and then generate a function tag "play song" according to the corresponding verb and noun, and bind the function tag of "play song" to the C model.

[0098] In the specific implementation, the corresponding function label is generated by the successful execution result of each voice model, and the function label is bound to the voice model. When the model is switched subsequently, the function label corresponding to each voice model can be directly obtained, and the intention supported by each voice model can be determined based on each function label.

[0099] Step S02: constructing a preset mapping relationship according to the support intentions corresponding to each of the speech models.

[0100] After determining the intents supported by each voice model, a preset mapping relationship can be constructed and saved based on each voice model and the supported intents corresponding to each voice model for subsequent direct query.

[0101] Accordingly, the step of determining the corresponding third speech model according to the target intention includes:

[0102] Step S21: Determine the third speech model corresponding to the target intention based on the preset mapping relationship.

[0103] Since the support intentions corresponding to each voice model have been determined, when determining the third voice model, the above-mentioned device can determine the corresponding support intention according to the target intention and select the corresponding third voice model according to the support intention.

[0104] This embodiment can predetermine the function label corresponding to each voice model, and generate a preset mapping relationship between the support intent corresponding to each function label and the voice model. Then, when determining the third voice model, the preset mapping relationship can be directly queried according to the target intent to determine the corresponding third voice model.

[0105] refer to Figure 4 , Figure 4 4 is a flow chart of the third embodiment of the in-vehicle voice interaction method of the present invention.

[0106] Considering that when there is only one voice model corresponding to a function label (for example, there is only one voice model in a music APP with the function label of "play music"), the above device can directly determine the corresponding third voice model according to the target intent. When there are multiple voice models with the same function label (for example, there are two music APPs at the same time, and the voice models therein are C model and D model respectively), in order to accurately determine which voice model can achieve the target intent, such as Figure 4 As shown, in this embodiment, the above step S21 includes:

[0107] Step S211: Determine a list of optional speech models corresponding to the target intent based on the preset mapping relationship.

[0108] It should be noted that since the D model is a voice model in another music APP, the function tag bound to the D model may also include "play song". When the above device determines the corresponding voice model according to the preset mapping relationship, it can filter out the C model and the D model from all the voice models, and generate a list of optional voice models for the C model and the D model.

[0109] Step S212: completing the target intent according to the analysis result.

[0110] It is understandable that since the target intent determined according to the function label is "play music", but when the current voice command is "play song c by singer b", the above-mentioned device can improve the target intent based on the analysis results of the semantic analysis of the current voice command, that is, improve "play music" to "play music, and play song c by singer b".

[0111] Furthermore, in order to accurately improve the target intent, in this embodiment, the step of determining the target intent corresponding to the current voice instruction according to the analysis result includes:

[0112] The primary intention and the secondary intention corresponding to the current voice command are determined according to the analysis results, and the primary intention is used as the target intention.

[0113] It should be understood that, continuing based on the above example, the above device can use the current voice command "play song c by singer b" as the main intention corresponding to "play music", the secondary intention to "play song c by singer b", and "play music" as the target intention.

[0114] It should be emphasized that the above-mentioned device can divide the extracted nouns into hyponyms and hyponyms, that is, the hypernym of "song c by singer b" is "music", and then the corresponding primary intention and secondary intention can be determined according to the determined hyponyms and hyponyms.

[0115] Correspondingly, the above step S212 includes: improving the target intention according to the secondary intention.

[0116] In a specific implementation, the above-mentioned device can combine the secondary intent with the main intent, and then improve the target intent, that is, combine "play song c by singer b" with "play music", and obtain the improved target intent of "play music, and play song c by singer b".

[0117] Step S213: Determine the third voice model from the optional voice model list through the refined target intent.

[0118] Since the present embodiment divides intents into primary intents and secondary intents, when determining the function labels of each voice model, the function labels can also be divided into primary function labels and sub-function labels, wherein the primary function labels correspond to the primary intents, and the sub-function labels correspond to the sub-function intents. Both the primary function labels and the sub-function labels can be determined based on the execution content executed by each voice model to be bound.

[0119] After determining the main function label and sub-function label bound to each voice model, the main support intention and sub-support intention supported by each voice model can be determined. For example, the main support intention of model C is determined to be "play music", and the sub-support intention is "play song c by singer b". The main support intention of model D is determined to be "play music", and the sub-support intention is "play song e by singer d". After perfecting the above target intentions, it can be determined that model C should be selected as the third voice model based on the perfected target intention.

[0120] Similarly, when the above-mentioned device is broadcasting, a voice reporting instruction can also be generated based on the improved target intention for broadcasting, so that the third voice model can accurately execute the current voice instruction.

[0121] This embodiment can divide the function labels into main function labels and sub-function labels to bind each voice model, and can also divide the main intention and the secondary intention, so as to accurately determine the third voice model.

[0122] In addition, an embodiment of the present invention further proposes a storage medium, on which an in-vehicle voice interaction program is stored. When the in-vehicle voice interaction program is executed by a processor, the steps of the in-vehicle voice interaction method described above are implemented.

[0123] Reference Figure 5 , Figure 5 This is a structural block diagram of the first embodiment of the vehicle system of the present invention.

[0124] like Figure 5 As shown, the vehicle system proposed in the embodiment of the present invention includes:

[0125] The semantic analysis module 501 is used to perform semantic analysis on the current voice instruction by using the first voice model when a current voice instruction is acquired, and the current voice instruction is a voice instruction that cannot be recognized by the second voice model;

[0126] A model determination module 502 is used to determine the target intention corresponding to the current voice instruction according to the analysis result, and determine the corresponding third voice model according to the target intention;

[0127] A wake-up word broadcasting module 503, used to broadcast a preset wake-up word corresponding to the third voice model, and the preset wake-up word is used to wake up the third voice model;

[0128] The instruction generation module 504 is used to generate a corresponding voice reporting instruction based on the target intention, and broadcast the voice reporting instruction so that the third voice model executes the voice reporting instruction.

[0129] The above-mentioned device of this embodiment can monitor the second voice model currently running, obtain the processing result of the second voice model processing the current voice instruction, and if the processing result is execution failure, the preset voice model can be awakened, and the current voice instruction can be input into the preset voice model, and the current voice instruction can be semantically analyzed by the preset voice model; the target intent corresponding to the current voice instruction is determined according to the analysis result, and the third voice model corresponding to the matching target application is determined based on the target intent; the preset wake-up word corresponding to the third voice model is determined and broadcasted; after the third voice model is awakened, the voice reporting instruction generated according to the target intent is broadcasted so that the third voice model executes the current voice instruction. Compared with the existing method that requires the user to wait for the current voice model to exit before waking up the target voice model, this embodiment can automatically wake up the target voice model according to the current voice instruction and broadcast the instruction without user operation, thereby simplifying the user operation and improving the user experience.

[0130] As an implementation mode, the instruction generation module 504 is also used to determine whether the latest voice instruction is received within a preset time period; if so, generate the latest reporting instruction according to the latest voice instruction, and broadcast the latest reporting instruction so that the third voice model executes the latest reporting instruction.

[0131] As an implementation mode, the instruction generation module 504 is further used to determine whether the latest voice instruction matches the third voice model; if so, an operation of generating a latest reporting instruction according to the latest voice instruction is executed.

[0132] Based on the above-mentioned first embodiment of the vehicle system of the present invention, a second embodiment of the vehicle system of the present invention is proposed.

[0133] In this embodiment, the semantic analysis module 501 is also used to obtain the function label corresponding to each speech model, and determine the support intention corresponding to each speech model based on each function label; and construct a preset mapping relationship according to the support intention corresponding to each speech model.

[0134] The model determination module 502 is further used to determine the third speech model corresponding to the target intent based on the preset mapping relationship.

[0135] As an implementation mode, the semantic analysis module 501 is also used to obtain the execution content corresponding to the voice model to be bound when the voice model to be bound successfully executes the received voice instruction; generate a corresponding function label based on the execution content, and bind the function label to the voice model to be bound.

[0136] Based on the above-mentioned embodiments of the vehicle system of the present invention, a third embodiment of the vehicle system of the present invention is proposed.

[0137] In this embodiment, the model determination module 502 is also used to determine a list of optional speech models corresponding to the target intent based on the preset mapping relationship; to improve the target intent according to the analysis result; and to determine the third speech model from the list of optional speech models through the improved target intent.

[0138] As an implementation manner, the model determination module 502 is also used to determine the primary intention and the secondary intention corresponding to the current voice instruction according to the analysis result, and use the primary intention as the target intention.

[0139] The model determination module 502 is further used to improve the target intent according to the secondary intent.

[0140] Other embodiments or specific implementations of the vehicle system of the present invention can refer to the above-mentioned method embodiments and will not be described in detail here.

[0141] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0142] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0143] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0144] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for in-vehicle voice interaction, characterized in that: The method is applied to a vehicle system, wherein the vehicle system has at least a first voice model, a second voice model, and a third voice model, and the method comprises the following steps: When a current voice instruction is acquired, semantic analysis is performed on the current voice instruction by using the first voice model, and the current voice instruction is a voice instruction that cannot be recognized by the second voice model; Determine the target intention corresponding to the current voice instruction according to the analysis result, and determine the corresponding third voice model according to the target intention; Broadcasting a preset wake-up word corresponding to the third voice model, where the preset wake-up word is used to wake up the third voice model; A corresponding voice reporting instruction is generated based on the target intention, and the voice reporting instruction is broadcast so that the third voice model executes the voice reporting instruction.

2. The in-vehicle voice interaction method according to claim 1, characterized in that: Before the step of performing semantic analysis on the current voice instruction by using the first voice model when the current voice instruction is acquired, the method further includes: Acquire a function label corresponding to each voice model, and determine a support intent corresponding to each voice model based on each function label; Constructing a preset mapping relationship according to the support intentions corresponding to each of the voice models; The step of determining the corresponding third speech model according to the target intention comprises: The third speech model corresponding to the target intent is determined based on the preset mapping relationship.

3. The in-vehicle voice interaction method according to claim 2, characterized in that: Before the step of obtaining the function labels corresponding to each speech model, the method further includes: When the voice model to be bound successfully executes the received voice instruction, obtaining the execution content corresponding to the voice model to be bound; Generate a corresponding function tag based on the execution content, and bind the function tag to the voice model to be bound.

4. The in-vehicle voice interaction method according to claim 2, characterized in that: The step of determining the third speech model corresponding to the target intention based on the preset mapping relationship includes: Determine a list of optional voice models corresponding to the target intent based on the preset mapping relationship; Performing intention improvement on the target intention according to the analysis result; The third voice model is determined from the list of optional voice models through the refined target intent.

5. The in-vehicle voice interaction method according to claim 4, characterized in that: The step of determining the target intention corresponding to the current voice instruction according to the analysis result includes: Determine the primary intention and the secondary intention corresponding to the current voice command according to the analysis result, and use the primary intention as the target intention; The step of perfecting the target intention according to the analysis result comprises: The target intent is refined according to the secondary intent.

6. The in-vehicle voice interaction method according to any one of claims 1 to 5, characterized in that: Before the step of generating a corresponding voice reporting instruction based on the target intention, the method further includes: Determining whether a latest voice command is received within a preset time period; If so, a latest reporting instruction is generated according to the latest voice instruction, and the latest reporting instruction is broadcasted so that the third voice model executes the latest reporting instruction.

7. The in-vehicle voice interaction method according to claim 6, characterized in that: Before the step of generating the latest reporting instruction according to the latest voice instruction, the method further includes: Determining whether the latest voice instruction matches the third voice model; If yes, the step of generating the latest reporting instruction according to the latest voice instruction is executed.

8. An in-vehicle voice interaction device, characterized in that: The device includes: a memory, a processor, and an in-vehicle voice interaction program stored in the memory and executable on the processor, wherein the in-vehicle voice interaction program is configured to implement the steps of the in-vehicle voice interaction method as described in any one of claims 1 to 7.

9. A storage medium, characterized in that: The storage medium stores an in-vehicle voice interaction program, and when the in-vehicle voice interaction program is executed by the processor, the steps of the in-vehicle voice interaction method as described in any one of claims 1 to 7 are implemented.

10. A vehicle computer system, characterized in that: The vehicle system comprises: A semantic analysis module, configured to perform semantic analysis on a current voice instruction by using a first voice model when a current voice instruction is acquired, wherein the current voice instruction is a voice instruction that cannot be recognized by the second voice model; A model determination module, used to determine the target intention corresponding to the current voice instruction according to the analysis result, and determine the corresponding third voice model according to the target intention; A wake-up word broadcasting module, used to broadcast a preset wake-up word corresponding to the third voice model, and the preset wake-up word is used to wake up the third voice model; An instruction generation module is used to generate a corresponding voice reporting instruction based on the target intention, and broadcast the voice reporting instruction so that the third voice model executes the voice reporting instruction.

Images