Vehicle machine voice recognition method and system

Abstract

The invention discloses a vehicle voice recognition method and system, and the method comprises the steps: obtaining a control voice, judging the identity of a user, and matching a personalized database and a free database based on the identity of the user; analyzing the control voice to form a control instruction, judging whether the control instruction is matched with data of the personalized database, and if so, outputting a final instruction in combination with the personalized database without customer response; if not, recording in a free database, calling a knowledge graph to assist search, and outputting a final instruction in cooperation with user response; the personalized database and the free database are generated based on the personality of the customer, the identity of the user is recognized through the processing module, then the data in the personalized database are associated, and the final instruction is output, so that the accuracy of the final instruction is improved, and the user experience is better.

Description

Technical Field

[0001] The present invention relates to the field of vehicle computer systems, and in particular to a speech recognition method and system for vehicle computer systems. Background Art

[0002] With the advancement of science and technology, the continuous iteration of computer intelligent algorithms, voice recognition technology has been continuously promoted from consumer electronics products to the automotive industry. The in-vehicle infotainment system integrates voice recognition systems to help drivers solve the problems of inconvenient manual operations during driving, such as making and receiving calls, navigation control, and adjusting the sunroof, etc., bringing faster and safer operation methods; Specifically, reasonable voice interaction to control the entertainment module of the car system can reduce the boredom of driving and allow drivers to concentrate more on driving. On the other hand, with the intelligent development of the car system, passengers also need to use voice to control the car system to play music, turn on lights, etc.; Under normal circumstances, according to the voice principle, the user's audio is input into the system, and the system needs to convert the audio file into text format, and then compare the text with the local or cloud database to determine the user's intention, and then control the execution, and broadcast the execution result in the form of voice; In the execution stage of the command, it is accompanied by feedback questions and answers from the car system to determine whether the action is correct, the output accuracy effect is average, and the interactive experience is average; Specifically, in terms of voice control of music playback, the versions performed by different singers are preferred by passengers with different personalities, so it is difficult to quickly find the expected song with existing voice operations; It can be seen that how to identify the user's commands individually, provide accurate retrieval capabilities, and output satisfactory expected instructions is an urgent problem to be solved by existing technologies. Summary of the invention

[0003] With regard to the prior art, the voice control interaction of the vehicle computer cannot output corresponding expected instructions according to the identities of the people in the vehicle, resulting in a poor interaction experience and low recognition accuracy. The present invention provides a solution.

[0004] To achieve the above object, the present invention provides a vehicle machine voice recognition method, comprising the following steps: Acquire control voice, determine user identity, and match the personality database and the free database based on the user identity; Parsing the control voice to form a control instruction, determining whether the control instruction matches the data in the personality database, and if so, outputting a final instruction in combination with the personality database without the need for a customer response; If not, it is recorded in the free database, and the knowledge graph is called to assist the search and the final instruction is output after cooperating with the user's response.

[0005] As an improved solution of the present application, before generating the control voice, it is also necessary to obtain a wake-up word, and only when the wake-up word appears, the subsequent voice is obtained.

[0006] As an improved solution of the present application, the user identity determination step and the personality database generation step are as follows: Pre-recording a user's voice, and constructing an identity model based on spectrum data generated by the user's voice; Generate multiple personality tags based on the identity model, collect the personality tags selected by the user to match and build the personality database; After acquiring the control voice, a verification model is constructed according to the control voice, and the verification model is matched with the identity model to determine the user identity.

[0007] As an improvement to the present application, after determining that the control instruction matches the data in the personality database, if the final instruction is rejected by the user within the expected time, a first sub-database associated with the personality database is generated, and the first sub-database collects multiple groups of control instructions. When the same instruction is collected next time, the control instruction with the closest matching time sorting is prioritized for output.

[0008] As an improved solution of the present application, when it is determined that the control instruction does not match the data in the personality database, a first sub-database associated with the free database is generated, and the first sub-database collects the control instruction records. When the number of times a control instruction is called out reaches a threshold, the control instruction is generated in the personality database.

[0009] The present application also provides a vehicle machine speech recognition system, comprising an acquisition module, a processing module and a database module connected to each other; The acquisition module acquires the control voice, the processing module determines the user identity, and matches the personality database and the free database generated by the database module based on the user identity; The processing module parses the control voice to form a control instruction, and determines whether the control instruction matches the data in the personality database. If so, it combines the personality database to output the final instruction without the need for customer response; if not, it records it in the free database, calls the knowledge graph to assist in the search, and outputs the final instruction after cooperating with the user's response.

[0010] As an improved solution of the present application, it also includes a wake-up module. Before generating the control voice, the wake-up module acquires a wake-up word, and acquires subsequent voices only after the wake-up word appears.

[0011] As an improved solution of the present application, the user identity determination step and the personality database generation step are as follows: Pre-recording a user's voice, and constructing an identity model based on spectrum data generated by the user's voice; Generate multiple personality tags based on the identity model, collect the personality tags selected by the user to match and build the personality database; After acquiring the control voice, a verification model is constructed according to the control voice, and the verification model is matched with the identity model to determine the user identity.

[0012] As an improvement to the present application, after the processing module determines that the control instruction matches the data in the personality database, if the final instruction is rejected by the user within the expected time, the database module generates a first sub-database associated with the personality database, and the first sub-database collects multiple groups of control instructions. When the same instructions are collected next time, the control instructions with the most recent matching time are prioritized for output.

[0013] As an improvement on the present application, when the processing module determines that the control instruction does not match the data in the personality database, the database module generates a second sub-database associated with the free database, and the second sub-database collects the control instruction records. When a control instruction is called out a number of times that reaches a threshold, the corresponding control instruction is generated in the personality database.

[0014] The beneficial effects of the present invention are as follows: compared with the prior art, the present invention provides a vehicle-mounted computer voice recognition method and system, which generates a personality database and a free database based on the customer's personality, first identifies the user's identity through a processing module, and then associates the data in the personality database to output the final instruction, thereby improving the accuracy of the final instruction and providing a better user experience; when the control instruction does not have biased data in the personality database, the user's control instruction is analyzed through a generalized knowledge graph, which will not affect the user's use of the vehicle-mounted computer system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0016] In order to more clearly illustrate the present invention, the present invention is further described below in conjunction with the accompanying drawings.

[0017] In the following description, the example details are given to provide a deeper understanding of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. It should be understood that the specific embodiments are only used to explain the present invention and are not used to limit the present invention.

[0018] It should be understood that when the terms "include" and / or "comprises" are used in this specification, they indicate the existence of the stated features, integers, steps, operations, elements or components, but do not exclude the existence or addition of one or more other features, integers, steps, operations, elements, components or combinations thereof.

[0019] In view of the technical problem that the voice control interaction of the vehicle machine cannot output the corresponding expected instructions according to the different identities of the people in the car, resulting in poor interaction experience and low recognition accuracy, in order to solve the above technical problems, this application provides a vehicle machine voice recognition method, please refer to the attached Figure 1 , including the following steps: Obtain control voice, determine user identity, and match the personality database and free database based on the user identity; Parse the control voice to form a control instruction, determine whether the control instruction matches the data in the personality database, and if so, combine the personality database to output the final instruction without the need for the customer to respond; If not, it will be recorded in the free database, and the knowledge graph will be called to assist the search and output the final instruction after cooperating with the user's response.

[0020] In this embodiment, a personalized database and a free database are generated based on the customer's personality. The user's identity is first identified through the processing module, and then the data in the personalized database is associated to output the final instruction, thereby improving the accuracy of the final instruction and providing a better user experience. When the control instruction does not have biased data in the personalized database, the user's control instruction is analyzed through a broad knowledge graph, which will not affect the user's use of the vehicle system.

[0021] In this embodiment, before generating the control voice, it is necessary to obtain the wake-up word, and the subsequent voice is obtained only after the wake-up word appears; in the process of vehicle interaction, by setting the wake-up word, it is convenient for the vehicle to distinguish whether the user is chatting with others or in voice interaction with the vehicle computer, which can reduce the generation of erroneous control actions.

[0022] For the identity recognition part, it can be judged by biometric recognition technology. There can be an identity sensing unit in the car, such as fingerprint, vein, iris recognition, etc. In order to better fit the driving situation, in this embodiment, it is judged by voice recognition and the corresponding hook personality database. Specifically, the user identity judgment steps and the personality database generation are as follows: Pre-record the user's voice and construct an identity model based on the spectrum data formed by the user's voice; Generate multiple personality tags based on the identity model, collect the personality tags selected by the user to match and build a personality database; After obtaining the control voice, a verification model is constructed based on the control voice, and the verification model is matched with the identity model to determine the user's identity; Each person has a different timbre, so they can form unique and high-frequency data. Therefore, the constructed identity models are highly unique. After matching the voices, the user's identity can be determined accordingly, and the identity recognition process can be completed when the user is awakened. Personality tags are media data commonly used by users. For example, if user A selects singer B as a personality tag, user A sends a voice request to the car music playback system to play track C. When the control instruction to play track C is obtained, the processing module preferentially considers the relationship between track C and singer B. If singer B has performed track C, singer B's version will be played first. In this embodiment, after determining that the control instruction matches the data in the personality database, if the final instruction is rejected by the user within the expected time, a first sub-database associated with the personality database is generated, and the first sub-database collects multiple groups of control instructions. When the same instruction is collected next time, the control instruction with the closest matching time order is prioritized for output; the first sub-database is used to modify the preference data formed by the personality tag. Continuing to use the above example to further illustrate: User A selects the preferred surface C in the personality tag, but the song C is highly popular, and there are versions performed by singers B, D, and E respectively. The processing module plays singer D first based on the on-demand popularity. version, but user A immediately finds that it is not the expected version, so he asks to switch. The processing module plays singer E's version according to the priority of on-demand popularity. If user A finds it, it will change again until singer B's version is played. At this time, the corresponding track C is associated with singer B, singer E and singer D respectively, and the above singer data are recorded in the first sub-database. When user A asks the car's music system to play track C again in the future, singer B's version will be played based on the playback time. Under this interactive effect, personality recognition can be performed based on the customer's body, so as to achieve a final command that meets the personality and expectations.

[0023] In this embodiment, when it is determined that the control instruction does not match the data in the personality database, a first sub-database associated with the free database is generated, and the first sub-database collects control instruction records. When the number of times a control instruction is called out reaches a threshold, the control instruction is generated in the personality database accordingly. The previous embodiment is used again to further illustrate: user A does not select curve C and singer B in the personality tag, and the control instruction given by user A plays curve C sung by singer B for 8-12 times, then the processing module associates singer B with curve C. Subsequently, user A only needs to issue a control instruction of "play singer B's song", and curve C will be played accordingly; or, user A issues a control instruction of "play curve C", and singer B's version will be played accordingly. By recording the user's occasional control instructions, the user's stage preferences can be recorded, and there will be no situation where repeated questions are asked to confirm, thereby achieving a final command that meets the personality and expectations, further improving the user experience.

[0024] The present application also provides a vehicle machine speech recognition system, comprising an acquisition module, a processing module and a database module connected to each other; The acquisition module acquires the control voice, the processing module determines the user identity, and matches the personality database and the free database generated by the database module based on the user identity; The processing module parses the control voice to form a control instruction, and determines whether the control instruction matches the data in the personality database. If so, it combines the personality database to output the final instruction without the need for customer response; if not, it is recorded in the free database, calls the knowledge graph to assist in the search, and outputs the final instruction after cooperating with the user's response.

[0025] It also includes a wake-up module. Before generating the control voice, the wake-up module acquires the wake-up word, and only acquires the subsequent voice after the wake-up word appears.

[0026] The steps for determining user identity and generating a personality database are as follows: Pre-record the user's voice and build an identity model based on the spectrum data formed by the user's voice; Based on the identity model, multiple personality tags are generated, and the personality tags selected by the user are collected to match and build a personality database; after obtaining the control voice, a verification model is built based on the control voice, and the verification model is matched with the identity model to determine the user's identity.

[0027] Specifically, after the processing module determines that the control instruction matches the data in the personality database, if the final instruction is rejected by the user within the expected time, the database module generates a first sub-database associated with the personality database. The first sub-database collects multiple groups of control instructions. When the same instruction is collected next time, the control instruction with the most recent matching time sorting is prioritized for output.

[0028] In another specific case, when the processing module determines that the control instruction does not match the data in the personality database, the database module generates a second sub-database associated with the free database. The second sub-database collects the control instruction records. When the number of times a control instruction is called out reaches a threshold, the control instruction is generated in the personality database.

[0029] In the above embodiment, the acquisition unit, the processing unit and the database module are used to execute and implement the above method steps, which can obtain the same beneficial effects as the above method, and therefore will not be described in detail.

[0030] The advantages of the present invention are: By generating a personalized database and a free database based on the customer's personality, the user's identity is first identified through the processing module, and then the final instruction is output after associating the data in the personalized database, thereby improving the accuracy of the final instruction and providing a better user experience; when the control instruction does not have biased data in the personalized database, the user's control instruction is analyzed through a broad knowledge graph, which will not affect the user's use of the vehicle system.

[0031] The above disclosures are only several specific embodiments of the present invention, but the present invention is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of the present invention.

Claims

1. A vehicle machine voice recognition method, characterized in that: The following steps are involved: Acquire control voice, determine user identity, and match the personality database and the free database based on the user identity; Parsing the control voice to form a control instruction, determining whether the control instruction matches the data in the personality database, and if so, outputting a final instruction in combination with the personality database without the need for a customer response; If not, it is recorded in the free database, and the knowledge graph is called to assist the search and the final instruction is output after cooperating with the user's response.

2. The vehicle machine voice recognition method according to claim 1, characterized in that: Before generating the control voice, the wake-up word needs to be acquired, and the subsequent voice is acquired only after the wake-up word appears.

3. The vehicle machine voice recognition method according to claim 1, characterized in that: The steps of determining the user identity and generating the personality database are as follows: Pre-recording a user's voice, and constructing an identity model based on spectrum data generated by the user's voice; Generate multiple personality tags based on the identity model, collect the personality tags selected by the user to match and build the personality database; After acquiring the control voice, a verification model is constructed according to the control voice, and the verification model is matched with the identity model to determine the user identity.

4. The vehicle machine voice recognition method according to claim 1, characterized in that: After determining that the control instruction matches the data in the personality database, if the final instruction is rejected by the user within the expected time, a first sub-database associated with the personality database is generated. The first sub-database collects multiple groups of control instructions. When the same instruction is collected next time, the control instruction with the most recent matching time is prioritized for output.

5. The vehicle machine voice recognition method according to claim 1, characterized in that: When it is determined that the control instruction does not match the data in the personality database, a first sub-database associated with the free database is generated, and the first sub-database collects the control instruction records. When the number of times a control instruction is called out reaches a threshold, the control instruction is generated in the personality database.

6. A vehicle voice recognition system, characterized in that: It includes an acquisition module, a processing module and a database module which are interconnected; The acquisition module acquires the control voice, the processing module determines the user identity, and matches the personality database and the free database generated by the database module based on the user identity; The processing module analyzes the control voice to form a control instruction, determines whether the control instruction matches the data in the personality database, and if so, outputs a final instruction in combination with the personality database without the need for a customer response; If not, it is recorded in the free database, and the knowledge graph is called to assist the search and the final instruction is output after cooperating with the user's response.

7. The vehicle-mounted voice recognition system according to claim 6, characterized in that: It also includes a wake-up module, which acquires a wake-up word before generating the control voice, and acquires subsequent voices only after the wake-up word appears.

8. The vehicle-mounted voice recognition system according to claim 6, characterized in that: The steps of determining the user identity and generating the personality database are as follows: Pre-recording a user's voice, and constructing an identity model based on spectrum data generated by the user's voice; Generate multiple personality tags based on the identity model, collect the personality tags selected by the user to match and build the personality database; After acquiring the control voice, a verification model is constructed according to the control voice, and the verification model is matched with the identity model to determine the user identity.

9. The vehicle-mounted voice recognition system according to claim 6, characterized in that: After the processing module determines that the control instruction matches the data in the personality database, if the final instruction is rejected by the user within the expected time, the database module generates a first sub-database associated with the personality database, and the first sub-database collects multiple groups of control instructions. When the same instruction is collected next time, the control instruction with the most recent matching time sorting is prioritized for output.

10. The vehicle-mounted voice recognition system according to claim 6, characterized in that: When the processing module determines that the control instruction does not match the data in the personality database, the database module generates a second sub-database associated with the free database, and the second sub-database collects the control instruction records. When the number of times a control instruction is called out reaches a threshold, the control instruction is generated in the personality database.

Images