Electronic device, voice recognition method, device and storage medium

Abstract

This invention discloses an electronic device, a speech recognition method, an apparatus, and a storage medium, relating to the field of speech recognition technology. The electronic device includes: a receiver configured to receive a voice command input by a user; and a controller coupled to the receiver, configured to: recognize the voice command in response to the voice command, obtaining multiple candidate recognition results corresponding to the voice command; determine the number of entities corresponding to the candidate recognition results based on the candidate recognition results; and, if the number of entities is less than or equal to a threshold number, analyze the scene information of the voice command using a scene decision module to determine the target recognition result of the voice command from the multiple candidate recognition results. Applying the technical solution of this invention can improve the accuracy of speech recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of speech recognition, and in particular, to an electronic device, a speech recognition method, a device, and a storage medium. Background Art

[0002] With the development of speech recognition technology, the application scenarios of speech interaction are becoming more and more common. For example, in application fields such as smart TVs, smart vehicles, smart homes, and smart robots, users can input speech commands through the speech function provided by the electronic device, so that the electronic device performs semantic analysis on the speech commands to identify the operation intention of the speech commands, and then controls the electronic device to perform corresponding operations according to the operation intention of the speech commands, such as running application programs, searching for content corresponding to the speech commands, etc.

[0003] However, due to the diversity of language types, the dialects in each region are different. For example, for some different characters, their pronunciations may be the same, such as "yue bing", and the corresponding characters are "military parade" and "moon cake". When performing recognition, sometimes the characters with the same pronunciation are recognized as incorrect characters or words. Therefore, the accuracy of speech recognition needs to be improved. Summary of the Invention

[0004] Embodiments of the present invention provide an electronic device, a speech recognition method, a device, and a storage medium, which are used to solve the problem of low accuracy of speech recognition in the prior art.

[0005] To achieve the above object, the embodiments of the present application adopt the following technical solutions:

[0006] According to one aspect of the embodiments of the present invention, an electronic device is provided, including: a receiver configured to receive a speech command input by a user; a controller coupled to the receiver, the controller being configured to: in response to the speech command, recognize the speech command to obtain a plurality of candidate recognition results corresponding to the speech command; determine the number of entities corresponding to the candidate recognition results; in the case where the number of entities is less than or equal to a number threshold, analyze the scene information of the speech command by using a scene decision module to determine a target recognition result of the speech command from the plurality of candidate recognition results.

[0007] In some embodiments, the controller is further configured to: in the case where the number of entities is greater than the number threshold, use a knowledge graph database to determine the relationship between the plurality of entities in each of the candidate recognition results; according to the relationship between the plurality of entities in each of the candidate recognition results, determine the candidate recognition result that meets the first output condition among the plurality of candidate recognition results as the target recognition result.

[0008] In some embodiments, the controller is further configured to: if the plurality of candidate recognition results do not meet the first output condition, then use the scene decision module to analyze the scene information of the voice command in order to determine the target recognition result of the voice command among the plurality of candidate recognition results.

[0009] In some embodiments, the scene information of the voice command includes at least one of a first scene information, a second scene information, and a third scene information; the scene decision module includes at least one of an environmental perception database, a spatiotemporal perception database, and a dialect perception database; the controller is configured to: match each of the candidate recognition results with information in the information database corresponding to the first scene information in the environmental perception database to obtain a first matching result for each of the candidate recognition results, wherein the first scene information includes device information and / or application information corresponding to the voice command; and / or match each of the candidate recognition results with information in the information database corresponding to the second scene information in the spatiotemporal perception database to obtain a second matching result for each of the candidate recognition results, wherein the second scene information includes time information and / or regional location information corresponding to the voice command; and / or match each of the candidate recognition results with information in the information database corresponding to the third scene information in the dialect perception database to obtain a third matching result for each of the candidate recognition results, wherein the third scene information includes the dialect information of the voice command; and, based on at least one of the first matching result, the second matching result, and the third matching result, determine the candidate recognition result that satisfies the second output condition among the plurality of candidate recognition results as the target recognition result.

[0010] In some embodiments, the controller is further configured to: when the plurality of candidate recognition results do not satisfy the second output condition, determine the candidate recognition result with the highest output probability as the target recognition result based on the output probability of each candidate recognition result.

[0011] In some embodiments, the controller is further configured to: determine the output probability difference between any two candidate recognition results among the plurality of candidate recognition results; if the minimum output probability difference among the plurality of output probability differences is greater than a difference threshold, determine the candidate recognition result with the highest output probability among the plurality of candidate recognition results as the target recognition result; if the maximum output probability difference among the plurality of output probability differences is less than or equal to the difference threshold, determine the number of entities corresponding to the candidate recognition result; if the number of entities is less than or equal to a quantity threshold, analyze the scene information of the voice command using a scene decision module to determine the target recognition result of the voice command among the plurality of candidate recognition results.

[0012] In some embodiments, the controller is further configured to: determine the operation intent corresponding to the voice command based on the target recognition result, so as to execute the control operation corresponding to the operation intent.

[0013] According to another aspect of the present invention, a speech recognition method is provided, applied to an electronic device, the method comprising: receiving a voice command input by a user; in response to the voice command, recognizing the voice command to obtain a plurality of candidate recognition results corresponding to the voice command; determining the number of entities corresponding to the candidate recognition results; and, if the number of entities is less than or equal to a number threshold, analyzing the scene information of the voice command using a scene decision module to determine the target recognition result of the voice command from the plurality of candidate recognition results.

[0014] In some embodiments, the method further includes: when the number of entities is greater than the number threshold, using a knowledge graph database to determine the relationship between multiple entities in each of the candidate recognition results; and determining the candidate recognition result that satisfies the first output condition among the multiple candidate recognition results as the target recognition result based on the relationship between multiple entities in each of the candidate recognition results.

[0015] In some embodiments, the method further includes: if the plurality of candidate recognition results do not meet the first output condition, then using the scene decision module to analyze the scene information of the voice command, so as to determine the target recognition result of the voice command among the plurality of candidate recognition results.

[0016] In some embodiments, the scene information of the voice command includes at least one of first scene information, second scene information, and third scene information; the scene decision module includes at least one of an environmental perception database, a spatiotemporal perception database, and a dialect perception database; the step of using the scene decision module to analyze the scene information of the voice command to determine the target recognition result of the voice command among the plurality of candidate recognition results includes: matching each of the candidate recognition results with information in the information database corresponding to the first scene information in the environmental perception database to obtain a first matching result for each of the candidate recognition results; the first scene information includes device information and / or application information corresponding to the voice command; and / or, matching each of the candidate... The recognition result is matched with the information in the information base corresponding to the second scene information in the spatiotemporal perception database to obtain a second matching result for each candidate recognition result, wherein the second scene information includes the time information and / or regional location information corresponding to the voice command; and / or, each candidate recognition result is matched with the information in the information base corresponding to the third scene information in the dialect perception database to obtain a third matching result for each candidate recognition result, wherein the third scene information includes the dialect information of the voice command; based on at least one of the first matching result, the second matching result, and the third matching result, the candidate recognition result that satisfies the second output condition among the plurality of candidate recognition results is determined as the target recognition result.

[0017] In some embodiments, the method further includes: if the plurality of candidate recognition results do not satisfy the second output condition, determining the candidate recognition result with the highest output probability as the target recognition result based on the output probability of each candidate recognition result.

[0018] In some embodiments, the method further includes: determining the output probability difference between any two candidate recognition results among the plurality of candidate recognition results; if the minimum output probability difference among the plurality of output probability differences is greater than a difference threshold, determining the candidate recognition result with the highest output probability among the plurality of candidate recognition results as the target recognition result; if the maximum output probability difference among the plurality of output probability differences is less than or equal to the difference threshold, determining the number of entities corresponding to the candidate recognition result; if the number of entities is less than or equal to a quantity threshold, using a scene decision module to analyze the scene information of the voice command, so as to determine the target recognition result of the voice command among the plurality of candidate recognition results.

[0019] In some embodiments, the method further includes: determining the operation intent corresponding to the voice command based on the target recognition result, so as to execute the control operation corresponding to the operation intent.

[0020] According to another aspect of the present invention, a voice recognition device is provided, applied to an electronic device, the device comprising: a receiving module for receiving a voice command input by a user; a first recognition module for recognizing the voice command in response to the voice command to obtain a plurality of candidate recognition results corresponding to the voice command; a determining module for determining the number of entities corresponding to the candidate recognition results; and a second recognition module for analyzing scene information of the voice command using a scene decision module when the number of entities is less than or equal to a number threshold, so as to determine the target recognition result of the voice command from the plurality of candidate recognition results.

[0021] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing at least one executable instruction, which, when executed on a display device, causes the display device to perform the operation of the speech recognition method as described above.

[0022] The electronic device, speech recognition method, apparatus, and storage medium provided in the embodiments of the present invention can respond to a user-inputted speech command, recognize the speech command, obtain multiple candidate recognition results corresponding to the speech command, determine the number of entities corresponding to the candidate recognition results, and, when the number of entities is less than or equal to a number threshold, use a scene decision module to analyze the scene information of the speech command to determine the target recognition result of the speech command.

[0023] By applying this solution, when multiple candidate recognition results corresponding to a voice command are obtained, and the number of entities in the candidate recognition results does not exceed the number threshold, the scene decision module can be used to analyze the scene information of the voice command, thereby clarifying and disambiguating each candidate recognition result, which can improve the accuracy of voice recognition. Attached Figure Description

[0024] Figure 1 This diagram illustrates the interaction between an electronic device and a control device according to an embodiment of the present invention.

[0025] Figure 2 A block diagram of the configuration of the control device in an embodiment of the present invention is shown;

[0026] Figure 3 A flowchart of a speech recognition method provided by an embodiment of the present invention is shown;

[0027] Figure 4 A structural diagram of a speech recognition system provided by an embodiment of the present invention is shown;

[0028] Figure 5 A flowchart of another speech recognition method provided by an embodiment of the present invention is shown;

[0029] Figure 6 A flowchart of another speech recognition method provided by an embodiment of the present invention is shown;

[0030] Figure 7 A schematic diagram of a knowledge graph provided by an embodiment of the present invention is shown;

[0031] Figure 8 A flowchart of another speech recognition method provided by an embodiment of the present invention is shown;

[0032] Figure 9 A flowchart of another speech recognition method provided by an embodiment of the present invention is shown;

[0033] Figure 10 A flowchart illustrating a voice command processing method provided by an embodiment of the present invention is shown.

[0034] Figure 11 A flowchart illustrating another voice command processing method provided by an embodiment of the present invention is shown;

[0035] Figure 12 A flowchart of another speech recognition method provided by an embodiment of the present invention is shown;

[0036] Figure 13 A schematic diagram of the structure of a speech recognition device provided in an embodiment of the present invention is shown. Detailed Implementation

[0037] To make the objectives and implementation methods of this application clearer, the exemplary implementation methods of this application will be clearly and completely described below with reference to the accompanying drawings of the exemplary embodiments of this application. Obviously, the exemplary embodiments described are only some embodiments of this application, and not all embodiments.

[0038] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0039] The terms "first," "second," "third," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar or related objects or entities, and do not necessarily imply a specific order or sequence, unless otherwise specified. It should be understood that such terms are interchangeable where appropriate.

[0040] The terms “comprising” and “having” and any variations thereof in this application are intended to cover but not exclude inclusion, for example, a product or device that includes a series of components is not necessarily limited to all the components that are clearly listed, but may include other components that are not clearly listed or that are inherent to such product or device.

[0041] This invention provides an electronic device and a voice recognition method. The voice recognition method can be applied to electronic devices, such as mobile phones, desktop computers, tablets, and wearable devices, that have voice recognition capabilities. Users can operate the electronic device via its voice function or through a control device. Figure 1 This diagram illustrates the interaction between an electronic device and a control device according to an embodiment of the present invention. Figure 1 As shown, a user can operate the electronic device 200 via a mobile terminal 300 or a control device 100. The control device 100 can be a remote control, and the remote control and the electronic device 200 can communicate via infrared protocol, Bluetooth protocol, or the remote control can also control the electronic device 200 wirelessly or via other wired means.

[0042] Users can control the electronic device 200 by inputting user commands through buttons on the remote control, voice input, and the control panel. For example, users can use the up and down buttons on the remote control to switch between displayed pages, the play / pause button to play or pause videos, and the voice input button to input voice commands to control the electronic device 200 to perform corresponding operations.

[0043] In some embodiments, users may also use mobile terminals, tablets, computers, laptops, and other smart devices to control electronic device 200. For example, users can control electronic device 200 through an application installed on the smart device, which, through configuration, can provide various controls to the user in an intuitive user interface on the screen associated with the smart device.

[0044] In some embodiments, the mobile terminal 300 can connect and communicate with the software application installed on the electronic device 200 via a network communication protocol to achieve one-to-one control operation and data communication. For example, a control command protocol can be established between the mobile terminal 300 and the electronic device 200, and a remote control keyboard can be synchronized to the mobile terminal 300. By controlling the user interface on the mobile terminal 300, the functions of controlling the electronic device 200 can be realized. Alternatively, the content displayed on the mobile terminal 300 can be transmitted to the electronic device 200 to achieve synchronous display.

[0045] like Figure 1As shown, electronic device 200 and server 400 can communicate via various methods, including local area network (LAN), wireless local area network (WLAN), and other networks. Server 400 can provide electronic device 200 with various content and interactive features. For example, electronic device 200 can interact by sending and receiving messages, using an electronic program guide (EPG), receiving software updates, or accessing a remotely stored digital media library. Server 400 can be a cluster or multiple clusters, and may include one or more types of servers.

[0046] Electronic device 200 can be a liquid crystal display (LCD), an organic light-emitting diode (OLED) display, a projection electronic device, or a smart terminal, such as a mobile phone, tablet computer, smart TV, laser projection device, or electronic table. Specific types, sizes, and resolutions of electronic devices are not limited.

[0047] Figure 2 A configuration block diagram of the control device 100 in an exemplary embodiment of the present invention is shown, as follows: Figure 2 As shown, the control device 100 includes a controller 110, a communication interface 130, a user input / output interface 140, a memory, and a power supply. The control device 100 can receive operation commands input by the user and convert the operation commands into commands that the electronic device 200 can recognize and respond to, thus acting as an intermediary between the user and the electronic device 200.

[0048] In some embodiments, the electronic device 200 can receive voice commands input by a user via a receiver. For example, when the receiver is a touch component, the touch component can form a touchscreen together with the display of the electronic device 200. On the touchscreen, the user can trigger the voice input function of the electronic device 200 by touching the screen to record voice commands.

[0049] In some embodiments, the receiver may also be an external control component, such as a mouse, remote control, or other device that establishes a communication connection with an electronic device. When the user performs different control operations on the external control component, the external control component can respond to the user's control operations by generating different control signals and sending the generated control signals to the controller 250.

[0050] For example, when the user presses the voice key on the remote control, the remote control can activate the voice input function. During the process of the user inputting a voice command, the remote control can synchronize the voice command to the display of the electronic device 200. At this time, the display can display a voice input identifier to indicate that the user is inputting a voice command.

[0051] In some embodiments, the receiver can also be a control component connected to the electronic device 200. Taking a desktop computer as an example, the control component can be a keyboard connected to the display. The user can input different control commands through the keyboard, such as click, switch, voice input and other operation commands.

[0052] Exemplarily, the user can input a voice command through a corresponding shortcut key. For example, when the user selects the "Ctrl" key and the "V" key, the controller 250 can receive a key signal, determine that the user has triggered a voice input operation. Then, the receiver can receive the voice command input by the user and control the electronic device 200 to perform corresponding operations according to the voice command.

[0053] Speech recognition technology is a technology in which a machine receives, recognizes, and understands a voice signal and converts it into a corresponding digital signal. Due to the vast territory of our country, there are various types of languages, and the languages used in each region are mostly dialects of their own regions. At the same time, with the wide application of the Internet, many homophonic words will derive different meanings. For example, for the pinyin "lu yao zhi ma li", the corresponding words can include "Lu Yao Zhi Ma Li" and "路遥知马力" (as the saying goes,路遥知马力, where "Lu Yao Zhi Ma Li" is the name of a certain TV drama).

[0054] In view of this situation, it is difficult to optimize the speech recognition model from the recognition engine end, resulting in the speech being easily recognized as incorrect text. Therefore, the accuracy of speech recognition is not high.

[0055] In view of the foregoing one or more problems, Figure 3 shows a flowchart of a speech recognition method provided by an embodiment of the present invention. This speech recognition method can be applied to an electronic device 200 as shown in Figure 1 The electronic device 200 can include a receiver and a controller coupled to the receiver.

[0056] Among them, the receiver can be used to receive the voice command input by the user, and the controller can analyze the voice command and control the electronic device to perform corresponding operations.

[0057] According to the speech recognition method provided in this embodiment, the electronic device can respond to the voice command input by the user, recognize the voice command, obtain multiple candidate recognition results corresponding to the voice command, determine the number of entities corresponding to the candidate recognition results, and when the number of entities is less than or equal to the number threshold, use the scene decision module to analyze the scene information of the voice command to determine the target recognition result of the voice command.

[0058] By applying this solution, when multiple candidate recognition results corresponding to a voice command are obtained, and the number of entities in the candidate recognition results does not exceed the number threshold, the scene decision module can be used to analyze the scene information of the voice command, thereby clarifying and disambiguating each candidate recognition result, which can improve the accuracy of voice recognition.

[0059] like Figure 3 As shown, the controller is configured to perform the following steps S310 to S340:

[0060] Step S310: Receive voice commands input by the user.

[0061] Voice commands are sound data acquired by the controller. These can be voice data input by the user through the voice input function of the electronic device, or voice data input by the user through external control components of the electronic device, such as a remote control or microphone. In some cases, voice commands may also be voice data acquired through other means, such as voice data downloaded from the network.

[0062] In some embodiments, voice commands may include scene information, which may include information about the electronic device and application when the voice command is received, the time information when the voice command is triggered, the location of the electronic device, etc.

[0063] Step S320: In response to the voice command, the voice command is recognized to obtain multiple candidate recognition results corresponding to the voice command.

[0064] Among them, the candidate identification result refers to the identification result of the candidate object.

[0065] When a voice command is received from a user, the controller can recognize the voice command. For example, it can use methods such as Hidden Markov Model, Artificial Neural Network, Support Vector Machine, and Machine Learning to recognize the voice command and obtain multiple candidate recognition results corresponding to the voice command.

[0066] For example, Figure 4 A structural diagram of a speech recognition system provided by an embodiment of the present invention is shown, as follows: Figure 4As shown, the speech recognition system 400 mainly consists of four components: feature extraction and signal processing, acoustic model, language model, and decoding search.

[0067] Feature extraction and signal processing are mainly used to enhance speech by eliminating noise and channel distortion, transforming the signal from the time domain to the frequency domain, and extracting suitable and representative feature vectors for subsequent acoustic models, taking the audio signal of the input speech command as input.

[0068] The acoustic model is the underlying model of the speech recognition system 400. Its purpose is to calculate the distance between the feature vector sequence of speech and each pronunciation template. In other words, the smaller the difference between the extracted speech features and a certain pronunciation, the more likely it is to be that pronunciation. Specifically, the acoustic model integrates acoustic and phonetics knowledge, taking the features generated by the feature extraction part as input, and generating an acoustic model score for the variable-length feature sequence.

[0069] Language models can learn the relationships between words by training on corpora, thus estimating the probability of hypothetical word sequences; this is also known as a language model score. Decoding search combines the acoustic model score and the language model score, selecting word sequences with higher overall output probabilities as candidate recognition results.

[0070] In some embodiments, word sequences with output probabilities higher than a minimum probability threshold can be identified as candidate recognition results. That is, among the word sequences corresponding to the voice commands determined by the speech recognition system, each word sequence with an output probability higher than the minimum probability threshold can be identified as a candidate recognition result. This can filter out some word sequences with excessively low output probabilities, reducing the amount of data for subsequent analysis.

[0071] By recognizing voice commands and identifying multiple candidate recognition results corresponding to the voice commands, multiple text recognition results corresponding to the voice commands can be determined, thus achieving preliminary recognition of voice commands.

[0072] Step S330: Determine the number of entities corresponding to the candidate recognition results.

[0073] An entity refers to a word or phrase in the candidate recognition result that represents a specific object, person, place, time, or other concrete entity. For example, in the candidate recognition result "I'm going to Hong Kong tomorrow," "tomorrow" and "Hong Kong" are entities, representing time and place respectively. Another example is the candidate recognition result "watching a movie," which contains only the entity word "movie," while the candidate recognition result "Zhang San's movie" contains both "Zhang San" and "movie."

[0074] When multiple candidate recognition results are obtained, the controller can determine the number of entities in each candidate recognition result, thereby obtaining the total number of entities in these candidate recognition results. Generally, the number of entities in each candidate recognition result is equal. For example, when the number of entities in multiple candidate recognition results is different, the value with the largest number of entities can be determined as the number of entities corresponding to the candidate recognition result.

[0075] Step S340: When the number of entities is less than or equal to the number threshold, the scene decision module is used to analyze the scene information of the voice command in order to determine the target recognition result of the voice command from multiple candidate recognition results.

[0076] The quantity threshold can be set to a default value, such as 1, or it can be customized by the developers. The target recognition result refers to the final recognition result of the voice command, and the target recognition result is one of multiple candidate recognition results. The scene decision module analyzes the scene and environment in which the voice command occurs. It can be used to analyze the user's geographical location, time information, spatial information, environment, and the device type and application that triggered the voice command when it occurs, in order to achieve optimal recognition performance.

[0077] When the number of entities is less than or equal to the number threshold, it indicates that there are few specific entity words in the candidate recognition results. In order to determine the target recognition result from the candidate recognition results, the scene decision module can be used to analyze the scene, environment, time, space and other information where the voice command occurs, so as to determine the target recognition result of the voice command from multiple candidate recognition results.

[0078] Using the above method, the scene decision module can be used to analyze the scene in which the voice command occurs, thereby clarifying and disambiguating each candidate recognition result and improving the accuracy of voice recognition.

[0079] In some embodiments, each candidate recognition result can be output as an array, where each entity in the candidate recognition result can include the entity name, the entity's pinyin, and the output probability. Table 1 below shows an example of a candidate recognition result, as follows:

[0080] Table 1

[0081] voice commands Candidate identification results yue bing [{yuebing:Military Parade, 0.98}, {yuebing:Mooncake, 0.97}] shu nian [{shunian: Year of Mature, 0.97}, {shunian: Year of Rat, 0.98}] zhang san de qi zi li si Zhang San's wife [{lisi:Li Si,0.95},{lisi:Li Si,0.94}]

[0082] "Mature Years" is the title of a certain TV series.

[0083] When the number of candidate recognition results is large, clarifying and disambiguating each candidate result increases the recognition time of voice commands. Therefore, in order to improve the efficiency of voice recognition, in some embodiments, reference is made to... Figure 5 As shown, the controller can also execute the following methods:

[0084] Step S510: Determine the output probability difference between any two candidate recognition results among multiple candidate recognition results.

[0085] For example, the output probability difference between any two candidate recognition results can be calculated among multiple candidate recognition results, that is, the absolute value of the difference between the output probabilities of two candidate recognition results can be calculated.

[0086] Step S520: If the minimum output probability difference among multiple output probability differences is greater than the difference threshold, the candidate recognition result with the highest output probability among multiple candidate recognition results is determined as the target recognition result.

[0087] The difference threshold can be customized by the developer or set to a default value, such as 0.05 or 0.1. Among multiple output probability differences, if the minimum output probability difference is greater than the difference threshold, it indicates a large difference between the output probability differences, with an even greater difference between the maximum and minimum output probabilities. In this case, the candidate recognition result with the highest output probability among the multiple candidate recognition results can be directly determined as the target recognition result.

[0088] For example, suppose there are three candidate recognition results. The output probability of the first candidate recognition result is 0.96, the output probability of the second candidate recognition result is 0.83, and the output probability of the third candidate recognition result is 0.72. The difference threshold is 0.05. After calculation, the output probability differences between the above three output probabilities are 0.13, 0.09, and 0.24, respectively. The minimum output probability difference of 0.09 is less than the difference threshold of 0.05. The difference between any two candidate recognition results is large. Therefore, the first candidate recognition result with an output probability of 0.96 can be directly determined as the target recognition result.

[0089] Step S530: If the maximum output probability difference among multiple output probability differences is less than or equal to the difference threshold, determine the number of entities corresponding to the candidate recognition result. If the number of entities is less than or equal to the number threshold, use the scene decision module to analyze the scene information of the voice command to determine the target recognition result of the voice command among multiple candidate recognition results.

[0090] Conversely, if the largest output probability difference is less than or equal to the difference threshold among multiple output probability differences, it indicates that the differences between the output probability differences are not significant, meaning that the output probability differences between the candidate recognition results are not significant. In this case, the number of entities corresponding to the candidate recognition results can be determined. Thus, when the number of entities is less than or equal to the number threshold, the scene decision module can be used to analyze the scene information of the voice command to determine the target recognition result from multiple candidate recognition results, thereby achieving the purpose of further recognizing the voice command.

[0091] By using the above method, it is possible to determine whether the scene decision module is needed to analyze the scene of the voice command based on the output probability differences between the candidate recognition results, so as to further clarify and disambiguate the candidate recognition results and thus determine the target recognition result, thereby improving the efficiency of voice analysis.

[0092] In some embodiments, reference Figure 6 As shown, the controller can also execute the following methods:

[0093] Step S610: When the number of entities exceeds the threshold, use the knowledge graph database to determine the relationships between multiple entities in each candidate recognition result.

[0094] A knowledge graph is a semantic network that represents the relationships between entities. In this embodiment, the knowledge graph database can be a knowledge graph about homophones built by developers, which can be used for disambiguation and decision-making regarding homophones in speech recognition. The knowledge graph database can include multiple pieces of knowledge, each of which can be represented as the following triple: [subject: topic] → predicate → [object].

[0095] When the number of entities exceeds a threshold, the candidate identification results can be clarified and disambiguated based on the relationships between entities. That is, the knowledge graph database is used to determine the relationships between multiple entities in each candidate identification result.

[0096] Figure 7 The illustration shows a schematic diagram of a knowledge graph provided by an embodiment of the present invention, such as... Figure 7 As shown in the knowledge graph, "movie" has entity relationships with "actor", "director" and "song", and "actor" acts in "movie", "director" directs "movie", the theme song of "movie" is "song", and the singer of "song" is "singer".

[0097] As can be seen, the relationships between entities are intricate and complex. Knowledge graphs can be used to express the connections and relationships between entities, intuitively demonstrating whether two entities are directly related and the nature of their relationship. Knowledge graph databases contain a large amount of data on relationships between entities, which can be used to analyze the correct relationships between entities in candidate identification results.

[0098] Taking the director of the movie "The Truth" as Zhao, we can find entity association data between the movie "The Truth" and "Zhao" using a knowledge graph database, but we cannot find entity association data between the movie "The Truth" and "Li". Therefore, we can determine that there is a relationship between the movie "The Truth" and "Zhao".

[0099] Step S620: Based on the relationship between multiple entities in each candidate recognition result, determine the candidate recognition result that satisfies the first output condition among the multiple candidate recognition results as the target recognition result.

[0100] The first output condition refers to the condition for selecting the target recognition result from multiple candidate recognition results. This condition can be that the candidate recognition result has a correct entity relationship. In other words, when determining the candidate recognition results with correct entity relationships among the candidate recognition results based on the knowledge graph database, it can be determined that the candidate recognition results with correct entity relationships satisfy the first output condition.

[0101] For example, in the above example, if the candidate recognition result is "I want to watch Zhao's movie 'The Truth'", then the candidate recognition result can be determined as the target recognition result.

[0102] For example, among multiple candidate recognition results "Searching for the idiom story 'Lu Yao Zhi Ma Li'" and "Searching for the idiom story 'Lu Yao Zhi Ma Li'", if the correct relationship between the entities can be determined based on the knowledge graph database as "Idiom story: Lu Yao Zhi Ma Li", then the candidate recognition result "Searching for the idiom story 'Lu Yao Zhi Ma Li'" can be determined as the final target recognition result.

[0103] Using the above method, when the number of entities exceeds a threshold, the relationship between entities can be checked and verified using a knowledge graph database, thereby selecting the target recognition result with the correct entity relationship from among the candidate recognition results.

[0104] Since knowledge graph databases may have imperfections, if the target recognition result cannot be determined from multiple candidate recognition results using a knowledge graph database, in some embodiments, the controller may also perform the following methods:

[0105] If multiple candidate recognition results do not meet the first output condition, the scene decision module is used to analyze the scene information of the voice command in order to determine the target recognition result of the voice command among the multiple candidate recognition results.

[0106] In other words, when the correct entity relationship cannot be determined using the knowledge graph database, and therefore the target recognition result cannot be determined, the controller can continue to use the scene decision module to analyze the scene information of the voice command and determine the target recognition result that matches the scene corresponding to the voice command among the candidate recognition results.

[0107] In some embodiments, the scene information of the voice command may include at least one of the first scene information, the second scene information, and the third scene information, and the scene decision module may include at least one of the environmental perception database, the spatiotemporal perception database, and the dialect perception database.

[0108] The environmental awareness database stores entity dictionaries corresponding to the first scene information. For example, it can store device information entity dictionaries and application entity dictionaries. The device information entity dictionaries can include entity dictionaries corresponding to different device types, and the application entity dictionaries can include entity dictionaries corresponding to different applications. The spatiotemporal awareness database stores entity dictionaries corresponding to the second scene information. For example, it can store time information entity dictionaries, including entity dictionaries for weekdays, rest days, and holidays, and it can also store entity dictionaries corresponding to regional locations, such as entity dictionaries for the southeast region or entity dictionaries for a specific city.

[0109] The dialect perception database is used to store entity lexicons corresponding to third-scene information. It can include dialect pronunciation information from various regions, such as storing entity lexicons of dialects from various regions, which can be used to represent entity words corresponding to each pronunciation.

[0110] It should be understood that the environmental perception database, spatiotemporal perception database, and dialect perception database can be divided into lexicons according to the corresponding data dimensions, and the fineness of the lexicon division can also be flexibly set.

[0111] Therefore, for reference Figure 8 As shown, the controller can execute the following methods:

[0112] Step S810: Match each candidate recognition result with the information in the information database corresponding to the first scene information in the environmental perception database to obtain the first matching result of each candidate recognition result.

[0113] The first scene information may include device information and / or application information corresponding to the voice command. Device information can indicate the type of electronic device triggering the voice command, such as a television, smartphone, in-vehicle terminal, air conditioner, etc. Application information refers to the application that triggers the voice command, such as a video application, audio application, home appliance management program, etc. The first matching result indicates whether each candidate recognition result successfully matches the information in the information base corresponding to the first scene information in the environmental perception database.

[0114] In order to determine the hardware environment that triggers the voice command, the controller can determine the device type and / or application corresponding to the voice command based on the first scene information, and then match each candidate recognition result with the database corresponding to the device type and / or application in the environment perception database to determine the first matching result of each candidate recognition result.

[0115] For example, the controller can determine the device type triggering the voice command based on the device information corresponding to the voice command, and then match each candidate recognition result with a device entity dictionary corresponding to the device type in the environment perception database to determine whether there is an entity in each device entity dictionary that matches a candidate recognition result. Alternatively, it can determine the application triggering the voice command based on the application information corresponding to the voice command, and then match each candidate recognition result with an application entity dictionary corresponding to the application in the environment perception database to determine whether there is an entity in each application entity dictionary that matches a candidate recognition result. For instance, when determining that the device information corresponding to the voice command is a television device, the controller can match each candidate recognition result with entities in the television device entity dictionary. When a candidate recognition result matches any entity in the television device entity dictionary, the controller determines that the candidate recognition result has been successfully matched.

[0116] Step S820: Match each candidate recognition result with the information in the information database corresponding to the second scene information in the spatiotemporal perception database to obtain the second matching result of each candidate recognition result.

[0117] The second scenario information may include the time information and / or regional location information corresponding to the voice command. The time information refers to the time when the voice command is triggered, which can be used to determine whether the voice command is triggered on a weekday, rest day, holiday, etc.

[0118] For example, in order to determine the time and location when a voice command is triggered, the controller can determine the time and / or location corresponding to the voice command based on the second scene information, and then search in the corresponding time entity lexicon and / or location entity lexicon whether there is an entity that matches each candidate recognition result.

[0119] For example, when the time of voice triggering is determined to be a rest day, the entity words in the time entity word library corresponding to the rest day can be matched with each candidate recognition result. When it is determined that any entity word matches a candidate recognition result, it means that the candidate recognition result is successfully matched.

[0120] Step S830: Match each candidate recognition result with the information in the information database corresponding to the third scene information in the dialect perception database to obtain the third matching result of each candidate recognition result.

[0121] The third scenario information may include dialect information for voice commands. For example, the pronunciation tones vary in different regions of my country. For instance, the tone sandhi rules in Shaanxi dialect are: "The first tone changes to the third tone, the second tone remains unchanged, the third tone changes to the fourth tone, and the fourth tone changes to the first tone."

[0122] Based on the third-scene information corresponding to the voice command, the controller can determine the dialect type of the voice command. It then matches the entity words in the dialect perception database of the corresponding region that have the same pronunciation as the entity words in each candidate recognition result to determine whether each candidate recognition result is a successful match. If the pronunciation of a candidate recognition result is the same as any entity word in the dialect perception database, and the two entity words also have the same characters, then the candidate recognition result is considered a successful match.

[0123] Step S840: Based on at least one of the first matching result, the second matching result, and the third matching result, determine the candidate recognition result that satisfies the second output condition among the multiple candidate recognition results as the target recognition result.

[0124] The second output condition refers to the condition for determining the target recognition result based on the first matching result, the second matching result, and the third matching result. When any one or more matching results are successfully matched, it can be determined that each candidate recognition result satisfies the second output condition, and the successfully matched candidate recognition result is the target recognition result.

[0125] Using the above method, the speech context of voice commands can be identified based on the environmental perception database, the spatiotemporal perception database, and the dialect perception database, thereby selecting the corresponding target recognition result from among the candidate recognition results.

[0126] It should be noted that the execution order of the above steps S810 to S830 is not fixed, and all of the above steps S810 to S830 can be executed, or only some of them can be executed. This embodiment does not make specific limitations in this regard.

[0127] In some embodiments, if multiple candidate recognition results satisfying the second output condition are determined in step S840, the candidate recognition result with the highest output probability can be determined as the target recognition result based on the output probability of the candidate recognition result satisfying the second output condition.

[0128] In some embodiments, the controller may also perform the following methods:

[0129] If none of the above candidate recognition results meet the second output condition, the candidate recognition result with the highest output probability is determined as the target recognition result based on the output probability of each candidate recognition result.

[0130] For example, when it is determined that none of the candidate recognition results can be matched successfully, and multiple candidate recognition results do not meet the second output condition, the candidate recognition result corresponding to the highest output probability can be determined as the target recognition result based on the output probability of each candidate recognition result.

[0131] Figure 9 A flowchart of another speech recognition method provided by an embodiment of the present invention is shown, such as... Figure 9 As shown, the following steps may be included:

[0132] Step S901: Receive voice commands input by the user.

[0133] Step S902: In response to the voice command, the voice command is recognized to obtain multiple candidate recognition results corresponding to the voice command.

[0134] Step S903: Determine whether the candidate identification result is a single entity. If not, proceed to step S904; if so, proceed to step S907.

[0135] If the candidate recognition result is not a single entity, it means that there are multiple entity words in the candidate recognition result. At this time, the candidate recognition result can be clarified and disambiguated according to the relationship between entity words. Therefore, step S904 can be executed to determine the relationship between multiple entities in each candidate recognition result using the knowledge graph database.

[0136] If the candidate recognition result is a single entity, it means that there is only one entity word in the candidate recognition result that represents a person, place, etc. In order to clarify and disambiguate the entity word, step S907 can be executed to match the first scene information of the voice command with the environmental perception database to obtain the first matching result of each candidate recognition result.

[0137] Step S904: Use a knowledge graph database to determine the relationships between multiple entities in each candidate recognition result.

[0138] Step S905: Determine whether each candidate recognition result meets the first output condition. If yes, proceed to step S906; otherwise, proceed to step S907.

[0139] The first output condition can be that there is a correct entity relationship in the candidate recognition results.

[0140] If a candidate recognition result satisfies the first output condition, it indicates that the candidate recognition result has a correct entity relationship. Therefore, step S906 can be executed to determine the candidate recognition result that satisfies the first output condition as the target recognition result. If none of the candidate recognition results satisfy the first output condition, it indicates that none of the candidate recognition results have a correct entity relationship. Therefore, step S907 can be executed to match the first scene information of the voice command with the environmental perception database to obtain the first matching result of each candidate recognition result.

[0141] Step S906: Determine the candidate recognition result that satisfies the first output condition from among the multiple candidate recognition results as the target recognition result.

[0142] Step S907: Match each candidate recognition result with the information in the information database corresponding to the first scene information in the environmental perception database to obtain the first matching result of each candidate recognition result.

[0143] The first scenario information may include device information and / or application information corresponding to the voice command.

[0144] For example, the device type triggering the voice command can be determined based on the device information corresponding to the voice command. Then, each candidate recognition result is matched with information in the environmental perception database and the device information database to determine whether there is information in the device information database that matches a candidate recognition result. If so, it means that the candidate recognition result has been successfully matched. Alternatively, the application triggering the voice command can be determined based on the application information corresponding to the voice command. Then, each candidate recognition result is matched with information in the environmental perception database and the application information database to determine whether there is information in the application information database that matches a candidate recognition result. If so, it means that the candidate recognition result has been successfully matched.

[0145] Step S908: Determine whether the first matching result is successful. If yes, proceed to step S914; otherwise, proceed to step S909.

[0146] When any candidate recognition result is determined to be a successful match based on the first matching result, step S914 can be executed to determine that the candidate recognition result meets the second output condition and is identified as the target recognition result. If it is determined that all candidate recognition results fail to match based on the first matching result, step S909 can be executed to continue analyzing the scene status of each candidate recognition result.

[0147] Step S909: Match each candidate recognition result with the information in the information database corresponding to the second scene information in the spatiotemporal perception database to obtain the second matching result of each candidate recognition result.

[0148] The second scene information may include the time information and / or regional location information corresponding to the voice command.

[0149] For example, when determining that the time triggering the voice command is the National Day holiday based on the time information corresponding to the voice command, each candidate recognition result can be matched with the information in the time information database corresponding to the National Day holiday in the spatiotemporal awareness database to determine whether there is information in the time information database corresponding to the National Day holiday that matches a certain candidate recognition result. If so, it means that the candidate recognition result has been successfully matched. Alternatively, the location of the region triggering the voice command can be determined based on the location information corresponding to the voice command, and then each candidate recognition result can be matched with the information in the location information database corresponding to the spatiotemporal awareness database to determine whether there is information in the location information database that matches a certain candidate recognition result. If so, it means that the candidate recognition result has been successfully matched.

[0150] Step S910: Determine whether the second matching result is successful. If yes, proceed to step S914; otherwise, proceed to step S911.

[0151] When any candidate recognition result is determined to be a successful match based on the second matching result, step S914 can be executed to determine that the candidate recognition result meets the second output condition and is identified as the target recognition result. If it is determined that all candidate recognition results fail to match based on the second matching result, step S11 can be executed to continue analyzing the scene conditions of each candidate recognition result.

[0152] Step S911: Match each candidate recognition result with the information in the information database corresponding to the third scene information in the dialect perception database to obtain the third matching result of each candidate recognition result.

[0153] The third scenario information may include dialect information for voice commands, such as a pronunciation entity dictionary for each region.

[0154] For example, based on the dialect type of the voice command, entity words in the dialect perception database of the corresponding region that have the same pronunciation as the entity words of each candidate recognition result can be matched to determine whether each candidate recognition result is a successful match.

[0155] Step S912: Determine whether the third matching result is a successful match. If yes, proceed to step S914; otherwise, proceed to step S913.

[0156] When any candidate recognition result is determined to be a successful match based on the third matching result, step S914 can be executed to determine that the candidate recognition result meets the second output condition and is thus identified as the target recognition result. If all candidate recognition results are determined to be unsuccessful matches based on the third matching result, step S13 can be executed to identify the candidate recognition result with the highest output probability as the target recognition result.

[0157] Step S913: Determine the candidate recognition result with the highest output probability as the target recognition result.

[0158] Step S914: Based on at least one of the first matching result, the second matching result, and the third matching result, determine the candidate recognition result that satisfies the second output condition among the multiple candidate recognition results as the target recognition result.

[0159] The second output condition refers to the condition for determining the target recognition result based on the first matching result, the second matching result, and the third matching result. When any one or more matching results are successfully matched, it can be determined that each candidate recognition result satisfies the second output condition, and the successfully matched candidate recognition result is the target recognition result.

[0160] After determining the target recognition result, in order to control the electronic device to output the feedback corresponding to the voice command, for example, referencing Figure 10 As shown, if the voice command is recognized as "play music '123'", the electronic device can respond to the voice command and output feedback voice, such as "Okay, I will play music '123' for you right away", and start playing music '123'.

[0161] In other words, in some embodiments, the controller can also perform the following methods:

[0162] The operation intent corresponding to the voice command is determined based on the target recognition result, and the control operation corresponding to the operation intent is executed.

[0163] For example, refer to Figure 11As shown, after a user inputs a voice command, the voice recognition method provided in this embodiment of the invention can be used to recognize the voice command, determine the target recognition result corresponding to the voice command, and then use natural language processing methods to perform semantic understanding on the target recognition result to determine the operation intent corresponding to the voice command. For example, the feedback operation corresponding to the target recognition result can be searched in the command library, thereby controlling the electronic device to execute the feedback operation and outputting the feedback command so that the user can know the progress of the electronic device's operation on the voice command.

[0164] Using the above method, the controller can receive voice commands input by the user, and based on the recognition of the voice commands, control the electronic device to output feedback on the voice commands, enabling the user to complete the purpose of voice control of the electronic device.

[0165] To further illustrate the speech recognition method provided in the embodiments of the present invention Figure 12 A flowchart of another speech recognition method provided by an embodiment of the present invention is shown, such as... Figure 12 As shown, the following steps may be included:

[0166] Step S1210: Receive voice commands input by the user.

[0167] Step S1220: Recognize the voice command to obtain the candidate recognition result corresponding to the voice command.

[0168] For example, a speech recognition system can be used to recognize speech commands, calculate the output probability of each word sequence corresponding to the speech command, and determine the word sequence with the higher output probability as the candidate recognition result.

[0169] Step S1230: Determine if multiple candidate recognition results exist. If yes, proceed to step S1240; otherwise, proceed to step S1260.

[0170] If multiple candidate recognition results exist, in order to determine the final target recognition result, step S1240 can be executed to input the multiple candidate recognition results into the scene decision module for processing. If only one candidate recognition result exists, it can be directly determined as the target recognition result, and step S1260 can be executed to input the target recognition result into the semantic understanding system to determine the operation intent corresponding to the voice command.

[0171] Step S1240: Input multiple candidate recognition results into the scene decision module for processing.

[0172] Specifically, this can be achieved using the embodiments of the present invention. Figure 9 The method shown processes multiple candidate recognition results; to avoid duplication, it will not be described in detail here.

[0173] Step S1250: Determine the target recognition result from among the candidate recognition results.

[0174] Based on the matching results output by the scene decision module, the target recognition result can be determined from each candidate recognition result.

[0175] Step S1260: Input the target recognition result into the semantic understanding system to determine the operation intent corresponding to the voice command.

[0176] Semantic understanding systems are a type of artificial intelligence technology designed to understand and interpret the meaning and intent within natural language text or speech input. Through lexical analysis, syntactic analysis, entity recognition, semantic parsing, and intent recognition, semantic understanding systems can convert target recognition results into computer-understandable and manipulable semantic representations.

[0177] After determining the target recognition result, the target recognition result can be input into the semantic understanding system to understand and interpret the meaning and operation intention of the voice command.

[0178] Step S1270: Execute the control operation corresponding to the operation intent.

[0179] After determining the operation intent corresponding to the voice command, the controller can execute the control operation corresponding to the operation intent, such as opening a page, opening or closing an application, or performing a settings function.

[0180] The above methods can be used to identify and analyze voice commands to determine the corresponding operational intent, thereby controlling electronic devices to perform control operations. This enables human-computer interaction in the form of voice commands and supports functions such as intelligent dialogue, information retrieval, knowledge Q&A, and device control.

[0181] In summary, the speech recognition method in this embodiment can recognize speech commands, determine the candidate recognition results corresponding to the speech commands, and analyze the scene information of the speech commands using the scene decision module based on the number of entities in the candidate recognition results, thereby clarifying and disambiguating each candidate recognition result and improving the accuracy of speech recognition.

[0182] This invention also provides a voice recognition device, such as... Figure 13As shown, the voice recognition device 1300 can be applied to electronic devices, and the voice recognition device 1300 may include: a receiving module 1310 for receiving voice commands input by a user; a first recognition module 1320 for recognizing the voice command in response to the voice command and obtaining multiple candidate recognition results corresponding to the voice command; a determining module 1330 for determining the number of entities corresponding to the candidate recognition results; and a second recognition module 1340 for analyzing the scene information of the voice command using a scene decision module when the number of entities is less than or equal to a number threshold, so as to determine the target recognition result of the voice command among multiple candidate recognition results.

[0183] In some embodiments, the second identification module 1340 can also be used to determine the relationship between multiple entities in each candidate identification result using a knowledge graph database when the number of entities is greater than a threshold; and to determine the candidate identification result that satisfies the first output condition among the multiple candidate identification results as the target identification result based on the relationship between multiple entities in each candidate identification result.

[0184] In some embodiments, the second recognition module 1340 can also be used to analyze the scene information of the voice command using the scene decision module if multiple candidate recognition results do not meet the first output condition, so as to determine the target recognition result of the voice command among multiple candidate recognition results.

[0185] In some embodiments, the scene information of the voice command includes at least one of first scene information, second scene information, and third scene information. The scene decision module includes at least one of an environmental perception database, a spatiotemporal perception database, and a dialect perception database. The second recognition module 1340 can also be used to match each candidate recognition result with the information in the information database corresponding to the first scene information in the environmental perception database to obtain a first matching result for each candidate recognition result. The first scene information includes device information and / or application information corresponding to the voice command. And / or, match each candidate recognition result with the information in the information database corresponding to the second scene information in the spatiotemporal perception database to obtain a second matching result for each candidate recognition result. The second scene information includes time information and / or regional location information corresponding to the voice command. And / or, match each candidate recognition result with the information in the information database corresponding to the third scene information in the dialect perception database to obtain a third matching result for each candidate recognition result. The third scene information includes the dialect information of the voice command. Based on at least one of the first matching result, second matching result, and third matching result, the candidate recognition result that satisfies the second output condition among the multiple candidate recognition results is determined as the target recognition result.

[0186] In some embodiments, the second identification module 1340 can also be used to determine the candidate identification result with the highest output probability as the target identification result when multiple candidate identification results do not meet the second output condition.

[0187] In some embodiments, the first recognition module 1320 can also be used to determine the output probability difference between any two candidate recognition results among multiple candidate recognition results; if the minimum output probability difference among multiple output probability differences is greater than the difference threshold, the candidate recognition result with the highest output probability among multiple candidate recognition results is determined as the target recognition result; if the maximum output probability difference among multiple output probability differences is less than or equal to the difference threshold, the number of entities corresponding to the candidate recognition result is determined; if the number of entities is less than or equal to the number threshold, the scene decision module is used to analyze the scene information of the voice command in order to determine the target recognition result of the voice command among multiple candidate recognition results.

[0188] In some embodiments, the second recognition module 1340 can also be used to determine the operation intent corresponding to the voice command based on the target recognition result, so as to execute the control operation corresponding to the operation intent.

[0189] Correspondingly, the specific details of each part of the above-mentioned voice recognition device have been described in detail in the above-mentioned electronic device implementation. For any undisclosed details, please refer to the implementation of the electronic device, and therefore will not be repeated here.

[0190] This invention provides a computer-readable storage medium storing at least one executable instruction that, when executed on an electronic device / speech recognition device, causes the electronic device / speech recognition device to perform the speech recognition method in any of the above method embodiments.

[0191] The executable instructions can be used to cause the electronic device / speech recognition device to perform the above-mentioned speech recognition method.

[0192] In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0193] More specific examples of computer-readable storage media (a non-exhaustive list) may include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0194] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Furthermore, the embodiments of this invention are not directed to any particular programming language.

[0195] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of the invention may be practiced without these specific details. Similarly, for the sake of brevity and to aid in understanding one or more aspects of the invention, in the description of exemplary embodiments of the invention above, various features of the embodiments are sometimes grouped together in a single embodiment, figure, or description thereof. The claims, which follow the detailed description, are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the invention.

[0196] Those skilled in the art will understand that the modules in the device of the embodiment can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiment can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components, except that at least some of such features and / or processes or units are mutually exclusive.

[0197] It should be noted that the above embodiments are illustrative of the invention and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The invention can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. An electronic device, characterized in that, The electronic device includes: The receiver is configured to receive voice commands input by the user; A controller coupled to the receiver is configured to: In response to the voice command, the voice command is recognized to obtain multiple candidate recognition results corresponding to the voice command; Determine the number of entities corresponding to the candidate recognition results, where the entities are words or phrases in the candidate recognition results that represent things, people, places, and times; When the number of entities is less than or equal to the number threshold, the scene decision module is used to analyze the scene information of the voice command in order to determine the target recognition result of the voice command from the multiple candidate recognition results.

2. The electronic device according to claim 1, characterized in that, The controller is also configured to: When the number of entities exceeds the number threshold, the relationship between multiple entities in each candidate recognition result is determined using a knowledge graph database; Based on the relationship between the multiple entities in each of the candidate recognition results, the candidate recognition result that satisfies the first output condition among the multiple candidate recognition results is determined as the target recognition result.

3. The electronic device according to claim 2, characterized in that, The controller is also configured to: If the multiple candidate recognition results do not meet the first output condition, the scene decision module is used to analyze the scene information of the voice command in order to determine the target recognition result of the voice command from the multiple candidate recognition results.

4. The electronic device according to any one of claims 1-3, characterized in that, The scene information of the voice command includes at least one of a first scene information, a second scene information, and a third scene information; the scene decision module includes at least one of an environmental perception database, a spatiotemporal perception database, and a dialect perception database; and the controller is configured to: Each of the candidate recognition results is matched with the information in the information database corresponding to the first scene information in the environmental perception database to obtain the first matching result of each of the candidate recognition results. The first scene information includes the device information and / or application information corresponding to the voice command. And / or, Each of the candidate recognition results is matched with the information in the information database corresponding to the second scene information in the spatiotemporal perception database to obtain a second matching result for each of the candidate recognition results. The second scene information includes the time information and / or regional location information corresponding to the voice command. And / or, The candidate recognition results are matched with the information in the information database corresponding to the third scene information in the dialect perception database to obtain the third matching result of each candidate recognition result. The third scene information includes the dialect information of the voice command. Based on at least one of the first matching result, the second matching result, and the third matching result, the candidate recognition result that satisfies the second output condition among the plurality of candidate recognition results is determined as the target recognition result.

5. The electronic device according to claim 4, characterized in that, The controller is also configured to: If the multiple candidate recognition results do not meet the second output condition, the candidate recognition result with the highest output probability is determined as the target recognition result based on the output probability of each candidate recognition result.

6. The electronic device according to any one of claims 1-3, characterized in that, The controller is also configured to: Determine the output probability difference between any two candidate recognition results among the plurality of candidate recognition results; If the smallest output probability difference among the plurality of output probability differences is greater than the difference threshold, the candidate recognition result with the highest output probability among the plurality of candidate recognition results shall be determined as the target recognition result; If the largest output probability difference among the multiple output probability differences is less than or equal to the difference threshold, the number of entities corresponding to the candidate recognition result is determined. If the number of entities is less than or equal to the number threshold, the scene decision module is used to analyze the scene information of the voice command in order to determine the target recognition result of the voice command among the multiple candidate recognition results.

7. The electronic device according to any one of claims 1-3, characterized in that, The controller is also configured to: The operation intent corresponding to the voice command is determined based on the target recognition result, so as to execute the control operation corresponding to the operation intent.

8. A speech recognition method, characterized in that, Applied to electronic devices, the method includes: Receive voice commands input by the user; In response to the voice command, the voice command is recognized to obtain multiple candidate recognition results corresponding to the voice command; Determine the number of entities corresponding to the candidate recognition results, where the entities are words or phrases in the candidate recognition results that represent things, people, places, and times; When the number of entities is less than or equal to the number threshold, the scene decision module is used to analyze the scene information of the voice command in order to determine the target recognition result of the voice command from the multiple candidate recognition results.

9. A voice recognition device, characterized in that, Applied to electronic devices, the device includes: The receiving module is used to receive voice commands input by the user; The first recognition module is used to respond to the voice command, recognize the voice command, and obtain multiple candidate recognition results corresponding to the voice command; The determining module is used to determine the number of entities corresponding to the candidate recognition results, wherein the entities are words or phrases representing things, people, places and times in the candidate recognition results; The second recognition module is used to analyze the scene information of the voice command using the scene decision module when the number of entities is less than or equal to the number threshold, so as to determine the target recognition result of the voice command from the multiple candidate recognition results.

10. A computer-readable storage medium, characterized in that, The storage medium stores at least one executable instruction, which, when executed on the display device, causes the display device to perform the operation of the speech recognition method as described in claim 8.

Images