Voice control method and device, computer device and computer readable storage medium

Abstract

The application provides a voice control method and device, computer equipment and a computer readable storage medium, comprising: obtaining a user audio signal; performing voice intent recognition on the user audio signal to obtain intent category information and keyword information; performing parameterization mapping processing on the intent category information and the keyword information to obtain processed intent category information and processed keyword information, which are used to control a target system. The method can effectively improve the voice recognition accuracy of the target system, thereby improving the control reliability of the target system.

Description

Technical Field

[0001] This application relates to the field of voice control technology, specifically to a voice control method, device, computer equipment, and computer-readable storage medium. Background Technology

[0002] Building Information Modeling (BIM) is a management information system for the construction industry. Based on intelligent 3D models, it enables engineering and construction professionals to gain relevant insights, analyses, and tools, thereby planning, designing, constructing, and managing buildings and infrastructure more efficiently. In smart parks, BIM systems can serve as the primary medium for displaying park information. Various park information can be integrated into the BIM system and displayed on a large screen within the park. The BIM system can act as an information platform for querying and controlling the park, achieving one-stop control and management.

[0003] Traditional control methods for smart park BIM systems mainly include keyboard and mouse control and touch screen control. However, due to drawbacks such as low selection efficiency and difficulty in cross-scene operation, they are gradually being replaced by smart park BIM systems incorporating voice recognition technology. However, while these voice control methods applicable to smart park BIM systems can solve problems such as difficult system operation, they still struggle to handle the complex technical terms within the park and BIM system. This often results in the inaccurate interpretation of the user's voice input intent, posing a challenge to the application of voice control technology in smart park BIM systems.

[0004] Therefore, existing voice control methods applicable to smart park BIM systems suffer from low voice recognition accuracy. Summary of the Invention

[0005] Therefore, it is necessary to provide a voice control method, device, computer equipment, and computer-readable storage medium to address the aforementioned technical problems, so as to improve the accuracy of voice recognition and enhance the control efficiency and robustness of the BIM system.

[0006] Firstly, this application provides a voice control method, including:

[0007] Acquire user audio signal;

[0008] Perform voice intent recognition on the user's audio signal to obtain intent category information and keyword information;

[0009] The intent category information and keyword information are parametrically mapped to obtain processed intent category information and processed keyword information, which are then used to control the target system.

[0010] In some embodiments of this application, the step of performing parameterized mapping processing on intent category information and keyword information to obtain processed intent category information and processed keyword information includes: performing parameterized mapping processing on intent category information to obtain a target control function among at least one pre-stored candidate control functions; performing parameterized mapping processing on keyword information to obtain a target word identifier among at least one pre-stored candidate word identifiers; using the target control function as the processed intent category information; and using the target word identifier as the processed keyword information.

[0011] In some embodiments of this application, the step of acquiring a user audio signal includes: acquiring an audio signal; performing noise reduction processing on the audio signal to obtain a noise-reduced audio signal; and filtering the echo signal in the noise-reduced audio signal based on a preset adaptive filtering algorithm to obtain the user audio signal.

[0012] In some embodiments of this application, the step of performing voice intent recognition on user audio signals to obtain intent category information and keyword information includes: inputting the user audio signal into a voice recognition model to obtain audio text information; obtaining valid text information corresponding to the audio text information based on a preset text mapping lookup table; and inputting the valid text information into an intent recognition model to obtain intent category information and keyword information.

[0013] In some embodiments of this application, the speech recognition model includes a first recognition model and a second recognition model. The step of inputting a user audio signal into the speech recognition model to obtain audio-text information includes: inputting the user audio signal into the first recognition model to obtain pinyin-text information; wherein, the first recognition model uses multiple pre-collected audios as training inputs and the audio pinyin obtained after converting the multiple audios into pinyin as the target output, and performs backpropagation training to obtain the audio-text information; inputting the pinyin-text information into the second recognition model to obtain audio-text information.

[0014] In some embodiments of this application, the first recognition model includes a convolutional module, a recurrent neural network module, and a fully connected module. The step of inputting a user audio signal into the first recognition model to obtain pinyin text information includes: inputting the user audio signal into the first recognition model; extracting audio features from the user audio signal through the convolutional module and obtaining the encoded features corresponding to the audio features; performing recurrent convolution sequence processing on the encoded features through the recurrent neural network module to obtain temporal features; and integrating the temporal features through the fully connected module to obtain pinyin text information.

[0015] Secondly, this application also provides a voice control method, including:

[0016] Receive audio signal acquisition trigger command;

[0017] In response to an audio signal acquisition trigger command, acquire the audio signal;

[0018] Based on the audio signal, obtain the processed intent category information and the processed keyword information;

[0019] The target system is controlled based on the processed intent category information and processed keyword information.

[0020] In some embodiments of this application, the step of acquiring an audio signal in response to an audio signal acquisition trigger command includes: in response to an audio signal acquisition operation command; if the audio signal acquisition operation command is a voice wake-up command, then acquiring a wake-up word, and after detecting that the wake-up word matches a preset valid wake-up word, acquiring the corresponding audio signal; if the audio signal acquisition operation command is a key input command, then acquiring the corresponding audio signal after detecting that a preset voice input key is triggered; if the audio signal acquisition operation command is a voice recognition command, then acquiring the audio frame to be recognized, and filtering out audio frames that do not contain silent frames as the corresponding audio signal.

[0021] In some embodiments of this application, the target system is controlled based on the processed intent category information and the processed keyword information, including: determining a target control function based on the processed intent category information; invoking the target control function and using the processed keyword information as the information input of the target control function to control the target system.

[0022] Thirdly, this application provides a voice control device, comprising:

[0023] The signal acquisition module is used to acquire the user's audio signal;

[0024] The intent recognition module is used to perform voice intent recognition on the user's audio signal to obtain intent category information and keyword information;

[0025] The parameter processing module is used to perform parameterized mapping processing on intent category information and keyword information to obtain processed intent category information and processed keyword information. The processed intent category information and processed keyword information are used to control the target system.

[0026] Fourthly, this application also provides a voice control device, comprising:

[0027] The instruction receiving module is used to receive audio signal acquisition trigger instructions;

[0028] The instruction response module is used to acquire audio signals in response to audio signal acquisition trigger instructions;

[0029] The signal transmission module is used to obtain processed intent category information and processed keyword information based on the audio signal;

[0030] The system control module is used to control the target system based on the processed intent category information and processed keyword information.

[0031] Fifthly, this application also provides a computer device, comprising:

[0032] One or more processors;

[0033] The memory; and one or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the processor to implement the voice control method.

[0034] In a seventh aspect, this application also provides a computer-readable storage medium having a computer program stored thereon, the computer program being loaded by a processor to execute the steps in the voice control method.

[0035] Eighthly, embodiments of this application provide a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the first and / or second aspects described above.

[0036] The aforementioned voice control method, apparatus, computer equipment, and computer-readable storage medium allow the server to obtain the intent category information and keyword information required for system control by recognizing the user's audio signal. Further parameterized mapping of the intent category information and keyword information yields system control commands, such as the processed intent category information and processed keyword information, thereby achieving precise control of the target system. The voice control method provided in this application, by recognizing the intent and keywords contained in the audio signal and performing parameterized mapping, enables the audio signal to be accurately matched to the control function used for system control. Compared to the traditional, singular method of using speech recognition keywords for system control, this application improves the accuracy of speech recognition and enhances the reliability of system control. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of a scenario for the voice control method in an embodiment of this application;

[0039] Figure 2 This is a flowchart illustrating the voice control method in an embodiment of this application;

[0040] Figure 3 This is a flowchart illustrating another voice control method in an embodiment of this application;

[0041] Figure 4 This is a schematic diagram of the structure of the voice control device in the embodiments of this application;

[0042] Figure 5 This is a schematic diagram of the structure of another voice control device in the embodiments of this application;

[0043] Figure 6 This is a schematic diagram of the structure of the computer device in the embodiments of this application. Detailed Implementation

[0044] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0046] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0047] This application provides a voice control method, apparatus, computer device, and computer-readable storage medium, which will be described in detail below.

[0048] See Figure 1 , Figure 1 This is a schematic diagram illustrating a scenario of the voice control method provided in this application, which can be applied to a voice control system. The voice control system includes a terminal 100 and a server 200. The terminal 100 can be a device that includes both receiving and transmitting hardware, i.e., a device with receiving and transmitting hardware capable of performing bidirectional communication over a bidirectional communication link. Such a device can include cellular or other communication devices, having a single-line display, a multi-line display, or a cellular or other communication device without a multi-line display. Specifically, the terminal 100 can be a desktop terminal or a mobile terminal, and can also be a mobile phone, tablet computer, laptop computer, etc. The server 200 can be a standalone server, or a server network or server cluster, including but not limited to computers, network hosts, single network servers, multiple network server sets, or cloud servers composed of multiple servers. The cloud server consists of a large number of computers or network servers based on cloud computing.

[0049] Those skilled in the art should understand that Figure 1 The application environment shown is merely one applicable scenario for the solution in this application and does not constitute a limitation on the application scenario of the solution in this application. Other application environments may include more than one. Figure 1 The number of computer devices shown is more or less, for example Figure 1 Only one server 200 is shown in the diagram. It is understood that the voice control system may also include one or more other servers or one or more other terminals, which are not specifically limited here. In addition, the voice control system may also include a memory for storing data, such as audio data.

[0050] Those skilled in the art should understand that Figure 1 The schematic diagram of the voice control system shown is merely an example. The voice control system and scenarios described in the embodiments of the present invention are intended to more clearly illustrate the technical solutions of the embodiments of the present invention and do not constitute a limitation on the technical solutions provided by the embodiments of the present invention. As those skilled in the art will know, with the evolution of voice control systems and the emergence of new business scenarios, the technical solutions provided by the embodiments of the present invention are also applicable to similar technical problems.

[0051] See Figure 2 This application provides a voice control method, which is mainly applied to the above-mentioned... Figure 1Taking server 200 as an example, the method includes steps S201 to S203, as follows:

[0052] S201, acquire user audio signal.

[0053] Among them, user audio signal can refer to information carrier that carries changes in the frequency and amplitude of user sound waves.

[0054] First, it should be noted that the server 200 used in this application embodiment can be a remote server independent of the voice recognition system, or it can be a server that forms a voice control system together with the terminal 100. The specific choice can be determined according to the actual application requirements. However, the following description will focus on saving local computing resources of the system and describe the voice control steps when the server 200 is a remote server. The target system to be controlled in this application embodiment can be a BIM system.

[0055] Specifically, before the server 200 obtains the control instructions of the BIM system, it first needs to obtain the user audio signal. The user audio signal can be obtained in the following ways: (1) from the terminal 100, which sends the collected user audio signal to the server 200; (2) from the same level server synchronously, that is, multiple servers and / or terminals can be used as blockchain nodes to form a blockchain system, such as a public blockchain system or a private blockchain system. Although the basic attributes of these two blockchain systems are different (the information stored in the public blockchain cannot be tampered with, while the information stored in the private blockchain can be tampered with), the same thing is that the audio signal stored at any node server can be requested and obtained by other node servers in the system; (3) from the upper-level server or from the lower-level server in a polling manner, that is, multiple servers have a hierarchical relationship. After the upper-level server updates the data, the lower-level server can request and obtain it in real time, and after the lower-level server updates the data, the upper-level server can poll and obtain it periodically.

[0056] It is understood that the selection of the above-mentioned public blockchain system, private blockchain system, request acquisition method or polling acquisition method can be determined according to the actual application requirements, and the embodiments of this application do not make specific limitations.

[0057] Furthermore, while the user audio signal acquisition methods provided in the above examples are specific to user audio signals, it is possible that the user audio signal is obtained by the server 200 processing the original signal, and the original signal is obtained from other devices through the three methods provided in the above examples. The user audio signal acquisition steps involved in this embodiment will be described in detail below.

[0058] In one embodiment, this step includes: acquiring an audio signal; performing noise reduction processing on the audio signal to obtain a noise-reduced audio signal; and filtering the echo signal in the noise-reduced audio signal based on a preset adaptive filtering algorithm to obtain a user audio signal.

[0059] Audio signals can refer to information carriers of regular sound wave frequency and amplitude changes that carry speech, music and / or sound effects.

[0060] Specifically, the solution described in this embodiment is the same as the solution mentioned in the previous embodiment: the user audio signal is obtained by server 200 based on the original signal processing, and the original signal is the audio signal mentioned in this embodiment. That is to say, before acquiring and analyzing the user audio signal, server 200 first obtains the audio signal from terminal 100. This audio signal can be sent by terminal 100 in real time, acquired by server 200 periodically, or obtained through periodic queries by a central server. After acquiring the audio signal, server 200 needs to preprocess the audio signal, including noise reduction and filtering, in order to perform speech recognition and reliably control the BIM system, to obtain the user audio signal required for subsequent analysis.

[0061] Furthermore, after the server 200 acquires the audio signal, in order to ensure the accuracy of speech recognition and the reliability of BIM system control, the audio signal must first be denoised to eliminate environmental noise and improve the signal-to-noise ratio. This can be achieved by using adaptive filtering algorithms such as Least Mean Square (LMS) and Recursive Least Squares Method (RLS), or by using the Recursive Least Squares based on QR decomposition (QR-RLS) algorithm, i.e., the square root adaptive filtering algorithm, to obtain the denoised audio signal.

[0062] Furthermore, after the server 200 acquires the denoised audio signal, echo cancellation is required to avoid echo interference affecting the speech recognition effect. For echo cancellation, this embodiment proposes to simulate the channel environment that generates the echo. Specifically, an adjustable filter is designed, and an adaptive algorithm (LMS algorithm) is used to adjust the filter's iterative update coefficients to estimate a desired signal that simulates the echo signal. This simulated echo is then subtracted from the denoised audio signal to achieve echo cancellation.

[0063] It should be noted that the adjustable parameters of the filter proposed in this embodiment are significant because they allow for processing and iterative learning of the input signal, enabling the filter to reach the desired value and achieve effective filtering. Specifically, the iterative learning of the filter first requires setting the objective function as the error signal, i.e., setting the objective as minimizing the error. Multiple iterations optimize the filter to obtain the optimal filter output, thus achieving echo cancellation.

[0064] S202, perform voice intent recognition on the user's audio signal to obtain intent category information and keyword information.

[0065] Among them, intent category information can refer to intent information applicable to BIM system control, which includes at least four categories: ① equipment control, ② page and model control, ③ subsystem query, and ④ navigation.

[0066] The keyword information can be slot keywords set for each intent category. A slot can refer to the conditions required for information retrieval. For example, "query the weather" requires two conditions: time and location, so "time" and "location" are called slots. The keywords involved in this embodiment also include four categories: ① control methods, locations, device names, etc., corresponding to "device control"; ② model names or page names, etc., corresponding to "page and model control"; ③ control methods, subsystem names, etc., corresponding to "subsystem query"; ④ destinations, etc., corresponding to "navigation".

[0067] Specifically, after acquiring the user's audio signal, the server 200 can use a trained neural network model to perform speech recognition on the user's audio signal, obtaining the speech recognition text corresponding to the user's audio signal. Then, intent recognition is performed on the speech recognition text to obtain interrelated intent category information and keyword information, thereby determining how the user wants to control the BIM system. For example, the user may want to control a specific device, building 3D information, BIM large screen page, subsystem information database, or navigation algorithm, and the keyword information can be used as the control object. The speech recognition and intent recognition steps involved in this embodiment will be described in detail below.

[0068] In one embodiment, this step includes: inputting the user's audio signal into a speech recognition model to obtain audio-text information; obtaining valid text information corresponding to the audio-text information based on a preset text mapping table; and inputting the valid text information into an intent recognition model to obtain intent category information and keyword information.

[0069] The speech recognition model can be a neural network model with a DeepSpeech2 structure, used to convert user audio signals into audio text information, which is the speech recognition text mentioned above.

[0070] Among them, the text mapping comparison table can be a comparison table containing incorrect texts and correct texts. The incorrect texts and correct texts are mapped in a one-to-one manner, and are used for the server 200 to check and fill in the gaps in the audio text information.

[0071] Among them, the intent recognition model can be a language representation model. For example, it can be a Bert (Bidirectional Encoder Representations from Transformer) model.

[0072] Specifically, the intent category information and keyword information that the server 200 needs to obtain are not directly obtained by analyzing through a certain algorithm or model, but are obtained by sequentially calling the speech recognition model, the text mapping comparison table, and the intent recognition model for analysis. Among them, the speech recognition model is used to convert the user audio signal into audio text information (speech recognition text), and obtain text translation information that can be further analyzed. This information could originally be input into the intent recognition model for intent recognition. However, since the keyword information (also known as slots) corresponding to each intent category information is a limited number of phrases, directly inputting it will lead to incorrect intent recognition. For example, "shenme" can be recognized as "什么" or "甚么", and so on. If the incorrect "甚么" is recognized in the previous step for the audio text information, it is very likely to cause an error in the analysis of the user's intent. Therefore, for this inevitable situation of incorrect speech recognition, the embodiment of this application proposes to use the text mapping comparison table to map the incorrect texts in the audio text information to the correct texts, and correct the audio text information to obtain valid text information.

[0073] Thus, the intent recognition model can classify the valid text information by intent, and find the slot keywords, obtaining the intent category information and keyword information, achieving an accurate mapping from speech to dedicated slots. Thereby, it solves the problem that the existing voice control technology is difficult to handle complex professional vocabulary in the BIM system, and also improves the situation where the user's voice control intent is difficult to be accurately determined, resulting in incorrect system control. Finally, it effectively improves the speech recognition accuracy and the system control reliability.

[0074] More specifically, the speech recognition steps performed by calling the speech recognition model will be explained in detail below. This section will focus on explaining the purpose of setting up the text mapping step: Considering that the slot keywords in practical applications are a finite number of word groups, and that there may be misidentifications after calling the speech recognition model, a secondary mapping step is proposed to improve the speech recognition effect, that is, to improve the accuracy of speech recognition. The so-called "secondary mapping" is to map the incorrect recognition results to the correct text. For example, the keyword "camera" is sometimes recognized as "camera system", so it is necessary to remap "camera system" to "camera" to initially improve the accuracy of audio-text information.

[0075] Furthermore, the BERT model takes valid text information as input and outputs intent category information and keyword information. For example, if the valid text information is "How to get to the sixth floor of Building 5 from here", the intent category information is the navigation category mentioned above (④), and the corresponding keyword information is the destination "the sixth floor of Building 5".

[0076] In one embodiment, the speech recognition model includes a first recognition model and a second recognition model. The step of inputting a user's audio signal into the speech recognition model to obtain audio-text information includes: inputting the user's audio signal into the first recognition model to obtain pinyin-text information; wherein the first recognition model uses multiple pre-collected audios as training inputs and the audio pinyin obtained after converting the multiple audios into pinyin as the target output, and performs backpropagation training to obtain the audio-text information; inputting the pinyin-text information into the second recognition model to obtain the audio-text information.

[0077] The first recognition model can be a sound model, which is mainly used to convert the user's audio signal into pinyin with tones, that is, to obtain pinyin text signals; the second recognition model can be a language model, which is mainly used to map pinyin to Chinese characters, that is, to obtain audio text information.

[0078] Specifically, server 200 uses the first and second recognition models to acquire audio-text information. First, the models need to be trained to enable analytical capabilities. This embodiment proposes that the data used to train the first recognition model consists of multiple audio pieces pre-collected using the Cartesian product algorithm, and these audio pieces are specifically collected based on the four types of intents described above. Simultaneously, the multiple audio pieces are augmented (also known as data augmentation or data expansion) through time warping, time masking, and spectral masking, thus slightly altering the audio and increasing the sample size.

[0079] For example, time warping can be used to randomly shift an image along a horizontal line, time masking can be used to eliminate a segment of audio, and spectral masking can be used to erase audio for a certain period of time.

[0080] More specifically, after the server 200 analyzes the user audio signal using the trained first recognition model and obtains the pinyin text information, it can then call the second recognition model to analyze the audio text information. In this embodiment, the second recognition model can adopt a Hidden Markov Model (HMM). The role of the HMM model is to make the phrase matching more reasonable.

[0081] For example, the text corresponding to the pinyin text information "shen2 me5" with tones is "什么". If a direct table mapping is done, incorrect recognition results such as "甚么" may be obtained. However, the HMM model will combine the previous one or two characters and calculate a probability. For example, if the first character is "什" and the second character is "me5", there is a higher probability of matching the character "么". Based on the probability analysis of the final result, the recognition accuracy of the audio text information can be improved, thereby improving the speech recognition accuracy.

[0082] In one embodiment, the first recognition model includes a convolutional module, a recurrent neural network module, and a fully connected module. The steps of inputting the user audio signal into the first recognition model to obtain the pinyin text information include: inputting the user audio signal into the first recognition model, extracting the audio features in the user audio signal through the convolutional module, and obtaining the encoded features corresponding to the audio features; performing a recurrent convolutional sequence processing on the encoded features through the recurrent neural network module to obtain the temporal features; integrating the temporal features through the fully connected module to obtain the pinyin text information.

[0083] Among them, the number of convolutional layers of the convolutional module can be "2", the number of recurrent neural network layers of the recurrent neural network module can be "3", and the number of fully connected layers of the fully connected module can be "1".

[0084] Specifically, after the server 200 inputs the user audio signal into the first recognition model, the convolutional module in the first recognition model first extracts the features of the user audio signal to obtain the audio features, and performs non-linear encoding on the audio features to obtain the encoded features. Since there is a high correlation between the input and output sequences of the speech signal, the recurrent neural network module performs a recurrent convolutional sequence processing on the encoded features, making the encoded features have temporal correlation and obtaining the temporal features. Finally, the temporal features are input into the fully connected module. The fully connected module integrates the encoded features from the previous two network modules and differentiates them, maps the features to the pinyin with tones, and obtains the pinyin text information to complete the speech recognition.

[0085] S203. Perform parametric mapping processing on the intent category information and the keyword information to obtain the processed intent category information and the processed keyword information. The processed intent category information and the processed keyword information are used to control the target system.

[0086] Parametric design is the process of writing engineering as functions, while the parametric mapping process involved in the embodiments of this application can be a query process from information conditions to target functions.

[0087] Specifically, after obtaining the intent category information and keyword information, the server 200 needs to further determine the program capable of executing the corresponding control in order to control the BIM system. These programs are pre-configured as control functions. The server 200 analyzes the intent category information and keyword information to determine the target control function, which then prompts the terminal 100 to call the target control function to execute the response control steps. Therefore, the server 200 needs to perform parameterized mapping processing on the intent category information and keyword information to obtain the information conditions required to lock the target control function, namely, the processed intent category information and processed keyword information, and then send these information conditions to the terminal 100.

[0088] In one embodiment, this step includes: performing parameterized mapping processing on intent category information to obtain a target control function among at least one pre-stored candidate control functions; performing parameterized mapping processing on keyword information to obtain a target word identifier among at least one pre-stored candidate word identifiers; using the target control function as the processed intent category information, and using the target word identifier as the processed keyword information.

[0089] Specifically, server 200 pre-stores at least one candidate control function and at least one candidate word identifier. Each candidate control function can be called to execute a control task, and each candidate word identifier corresponds to a keyword. Server 200 performs parameterized mapping processing on the acquired intent category information, thereby querying and determining the target control function from at least one candidate control function. Simultaneously, it performs parameterized mapping processing on the acquired keyword information, thereby querying and determining the target word identifier from at least one candidate word identifier. Finally, processed intent category information and processed keyword information are obtained.

[0090] For example, as mentioned above, this application involves four categories of intents, therefore, there are also four types of control functions. Since there are a large number of keywords, and different keywords apply to different intents, this application proposes using multiple mapping tables for parameterization, that is, binding keyword information to its globally unique identifier. For example, if the intent category information is "page redirection" and the corresponding keyword information is "homepage," then the control function function_two (UUID=0) can be called. This maps an audio content to the function. Terminal 100 locks the target control function and the target keyword identifier, and can then execute control steps that meet the user's needs.

[0091] Specifically, once the server 200 obtains the processed intent category information and the processed keyword information, it can send these two pieces of information to the terminal 100. After receiving the information, the terminal 100 can query and obtain the target control function and the target word identifier, and then use the target word identifier as the input of the target control function to execute the control command steps provided in the user's audio signal.

[0092] In addition to sending the processed intent category information and processed keyword information to the terminal 100, the server 200 can also use text-to-speech (TTS) technology to obtain the speech corresponding to the audio text information, and then feed the speech back to the terminal 100 for playback.

[0093] The voice control method described in the above embodiments allows the server to obtain the intent category information and keyword information required for system control by recognizing the user's audio signal. Further parameterized mapping of the intent category information and keyword information yields system control commands, such as the processed intent category information and processed keyword information, thereby achieving precise control of the target system. The voice control method provided in this application, by recognizing the intent and keywords contained in the audio signal and performing parameterized mapping, enables the audio signal to be accurately matched to the control function used for system control. Compared to the traditional, singular method of using speech recognition keywords for system control, this application improves the accuracy of speech recognition and enhances the reliability of system control.

[0094] See Figure 3 This application also provides another voice control method, which is mainly applied to the above-mentioned... Figure 1 Taking terminal 100 as an example, the method includes steps S301 to S304, as follows:

[0095] S301 receives audio signal acquisition trigger commands.

[0096] The audio signal acquisition trigger command can be any one of the following: a voice wake-up command, a key input command, or a voice recognition command.

[0097] Specifically, terminal 100 runs a BIM system and can control sub-modules of the BIM system, such as equipment, 3D building information, BIM large screen pages, subsystem information databases, and navigation algorithms. However, before responding to user commands and executing corresponding controls, it can collect the voice control audio issued by the user. This voice control audio can be collected when an audio signal collection trigger command is detected and received. Therefore, based on actual business needs, the audio signal collection trigger command can be set to any one of the following: a voice wake-up command, a key input command, or a voice recognition command, to trigger terminal 100 to enable the audio collection function.

[0098] S302, in response to the audio signal acquisition trigger command, acquires the audio signal.

[0099] Specifically, after receiving the audio signal acquisition trigger command, terminal 100 can respond to the command and acquire the audio signal that can instruct the BIM system to perform corresponding control. This audio signal is the voice control audio issued by the user mentioned above. The audio signal acquisition steps involved in this embodiment will be described in detail below.

[0100] In one embodiment, this step includes: responding to an audio signal acquisition operation command; if the audio signal acquisition operation command is a voice wake-up command, acquiring a wake-up word, and acquiring the corresponding audio signal after detecting that the wake-up word matches a preset valid wake-up word; if the audio signal acquisition operation command is a key input command, acquiring the corresponding audio signal after detecting that a preset voice input key is triggered; if the audio signal acquisition operation command is a voice recognition command, acquiring the audio frame to be recognized, and filtering out audio frames that do not contain silent frames as the corresponding audio signal.

[0101] Specifically, the terminal 100 may have at least one valid wake-up word pre-set, such as "Hello, campus". When it collects a wake-up word issued by the user and determines that the wake-up word matches a valid wake-up word, it can start collecting the user's subsequent voice as the audio signal required at the moment.

[0102] More specifically, the terminal 100 has an interactive interface that displays a virtual button. When the user presses and holds the virtual button, the terminal 100 can collect the user's voice while the button is pressed. Once the button is no longer pressed, the terminal 100 determines the currently collected voice as the audio signal required for subsequent processing.

[0103] Furthermore, the terminal 100 can collect sounds emitted by the surrounding environment in real time, divide the sound into audio frames, extract audio features, and then identify and classify speech frames and silence frames in the audio features. Finally, the speech frames are extracted as the audio signals required for subsequent processing.

[0104] For example, the peak portion of a sound wave curve can be identified as a speech frame, while the smooth portion can be identified as a silent frame. By extracting the sound wave from the peak portion, the desired audio signal can be obtained.

[0105] S303, based on the audio signal, obtains the processed intent category information and the processed keyword information.

[0106] Specifically, to save local resources and storage space, terminal 100 can send the collected audio signal to server 200, so that server 200 can obtain the user's audio based on the audio signal, and then obtain and feedback the processed intent category information and processed keyword information based on the user's audio signal. The information feedback steps involved in this embodiment have been described in detail above and will not be repeated here.

[0107] S304, based on the processed intent category information and processed keyword information, control the target system.

[0108] In one embodiment, after this step, the method further includes: determining a target control function based on the processed intent category information; invoking the target control function and using the processed keyword information as the information input of the target control function to control the target system.

[0109] Specifically, after receiving the processed intent category information and processed keyword information from the server 200, the terminal 100 analyzes the information in the same way as the server 200. That is, it determines the target control function based on the processed intent category information and the target word identifier based on the processed keyword information. Then, it calls the target control function to execute the control logic related to the target word identifier, which can meet the user's needs and execute the control of the target system (such as the BIM system) as instructed by the user. This includes, but is not limited to, controlling devices, pages, models, etc., as well as controlling information queries and navigation displays.

[0110] The voice control method described in the above embodiments allows the terminal to receive and respond to audio signal acquisition trigger commands, further acquire audio signals, and then use the audio signals to obtain information that can accurately control the target system. This not only saves local resources but also enhances the reliability and stability of audio analysis results, making the control accuracy of the target system higher.

[0111] To better implement the voice control method in the embodiments of this application, a voice control device is also provided in the embodiments of this application, such as... Figure 4 As shown, the voice control 400 for the target system includes:

[0112] Signal acquisition module 410 is used to acquire user audio signals;

[0113] The intent recognition module 420 is used to perform voice intent recognition on the user's audio signal to obtain intent category information and keyword information;

[0114] The parameter processing module 430 is used to perform parameterized mapping processing on the intent category information and keyword information to obtain the processed intent category information and processed keyword information. The processed intent category information and processed keyword information are used to control the target system.

[0115] In some embodiments of this application, the signal acquisition module 410 is further configured to acquire an audio signal; perform noise reduction processing on the audio signal to obtain a noise-reduced audio signal; and perform echo processing on the noise-reduced audio signal based on a preset adaptive filtering algorithm to obtain a user audio signal.

[0116] In some embodiments of this application, the intent recognition module 420 is further configured to input the user's audio signal into the speech recognition model to obtain audio text information; obtain the valid text information corresponding to the audio text information based on a preset text mapping table; and input the valid text information into the intent recognition model to obtain intent category information and keyword information.

[0117] In some embodiments of this application, the speech recognition model includes a first recognition model and a second recognition model. The intent recognition module 420 is further used to input the user's audio signal into the first recognition model to obtain pinyin text information. The first recognition model is trained by backpropagation using multiple pre-collected audios as training inputs and the audio pinyin obtained after converting the multiple audios into pinyin as the target output. The pinyin text information is input into the second recognition model to obtain audio text information.

[0118] In some embodiments of this application, the first recognition model includes a convolutional module, a recurrent neural network module, and a fully connected module. The intent recognition module 420 is further configured to input the user's audio signal into the first recognition model, extract audio features from the user's audio signal through the convolutional module, and obtain the encoded features corresponding to the audio features; perform recurrent convolutional sequence processing on the encoded features through the recurrent neural network module to obtain temporal features; and integrate the temporal features through the fully connected module to obtain the pinyin text information.

[0119] In some embodiments of this application, the parameter processing module 430 is further configured to perform parameterized mapping processing on the intent category information to obtain a target control function among at least one pre-stored candidate control functions; perform parameterized mapping processing on the keyword information to obtain a target word identifier among at least one pre-stored candidate word identifiers; use the target control function as the processed intent category information, and use the target word identifier as the processed keyword information.

[0120] In the above embodiments, the server can obtain the intent category information and keyword information required for system control by performing voice intent recognition on the user's audio signal. Further parameterized mapping processing of the intent category information and keyword information yields system control commands, such as the processed intent category information and processed keyword information, thereby achieving precise control of the target system. The voice control method provided in this application, by recognizing the intent and keywords contained in the audio signal and performing parameterized mapping processing on the intent and keywords, enables the audio signal to be accurately matched to the control function used for system control. Compared with the traditional single method of using voice recognition keywords for system control, this application improves the accuracy of voice recognition and enhances the reliability of system control.

[0121] To better implement the voice control method in the embodiments of this application, another voice control device is also provided in the embodiments of this application, such as... Figure 5 As shown, the voice control 500 for the target system includes:

[0122] The instruction receiving module 510 is used to receive audio signal acquisition trigger instructions;

[0123] Command response module 520 is used to acquire audio signals in response to audio signal acquisition trigger commands;

[0124] The signal transmission module 530 is used to obtain processed intent category information and processed keyword information based on the audio signal;

[0125] The system control module 540 is used to control the target system based on the processed intent category information and the processed keyword information.

[0126] In some embodiments of this application, the instruction response module 520 is further configured to respond to an audio signal acquisition operation instruction; if the audio signal acquisition operation instruction is a voice wake-up instruction, then the wake-up word is acquired, and after detecting that the wake-up word matches a preset valid wake-up word, the corresponding audio signal is acquired; if the audio signal acquisition operation instruction is a key input instruction, then the corresponding audio signal is acquired after detecting that a preset voice input key is triggered; if the audio signal acquisition operation instruction is a voice recognition instruction, then the audio frame to be recognized is acquired, and audio frames that do not contain silent frames are selected as the corresponding audio signal.

[0127] In some embodiments of this application, the system control module 540 is used to determine a target control function based on the processed intent category information; call the target control function and use the processed keyword information as the information input of the target control function to control the target system.

[0128] In the above embodiments, the terminal can further acquire audio signals by receiving and responding to audio signal acquisition trigger commands, and then use the audio signals to obtain information that can accurately control the target system. This not only saves local resources, but also enhances the reliability and stability of audio analysis results, making the control accuracy of the target system higher.

[0129] For specific limitations regarding the voice control device, please refer to the limitations on the voice control method above, which will not be repeated here. Each module in the aforementioned voice control device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independently of the processor in the computer device, or stored in software in the memory of the computer device, so that the processor can call and execute the operations corresponding to each module.

[0130] In some embodiments of this application, the voice control device 400 can be implemented as a computer program, which can be implemented in, for example... Figure 6 The computer device shown operates on this device. The computer device's memory can store the various program modules that make up the voice control device 400, for example, Figure 4 The signal acquisition module 410, intent recognition module 420, and parameter processing module 430 are shown. The computer program comprised of these modules causes the processor to execute the steps of the voice control methods in the various embodiments of this application described in this specification.

[0131] For example, Figure 6 The computer device shown can be used as follows Figure 4 The signal acquisition module 410 in the voice control device 400 shown executes step S201. The computer device can execute step S202 through the intent recognition module 420. The computer device can execute step S203 through the parameter processing module 430. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with external computer devices via a network connection. When the computer program is executed by the processor, it implements a voice control method.

[0132] Those skilled in the art will understand that Figure 6The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0133] In some embodiments of this application, a computer device is provided, including one or more processors; a memory; and one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processor as described in the voice control method. The steps of the voice control method here may be steps from the voice control methods of the various embodiments described above.

[0134] In some embodiments of this application, a computer-readable storage medium is provided, storing a computer program. The computer program is loaded by a processor, causing the processor to execute the steps of the aforementioned voice control method. The steps of the voice control method here can be the steps from the voice control methods of the various embodiments described above.

[0135] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0136] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0137] The above provides a detailed description of a voice control method, apparatus, computer device, and computer-readable storage medium provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A voice control method, characterized by, The method comprises the following steps: obtaining a user audio signal; performing speech intent recognition on the user audio signal to obtain intent category information and keyword information; performing parameterized mapping processing on the intent category information and the keyword information to obtain processed intent category information and processed keyword information, which are used to control a target system; the step of performing speech intent recognition on the user audio signal to obtain intent category information and keyword information comprises: inputting the user audio signal into a speech recognition model to obtain audio text information; based on a pre-set text mapping table, obtaining valid text information corresponding to the audio text information; inputting the valid text information into an intent recognition model to obtain the intent category information and the keyword information; the intent category information includes intent information for BIM system control; the keyword information includes slot keywords set for each intent category.

2. The method of claim 1, wherein, the step of performing parameterized mapping processing on the intent category information and the keyword information to obtain processed intent category information and processed keyword information comprises: performing parameterized mapping processing on the intent category information to obtain a target control function in at least one pre-stored candidate control function; performing parameterized mapping processing on the keyword information to obtain a target word identifier in at least one pre-stored candidate word identifier; the target control function is used as the processed intent category information, and the target word identifier is used as the processed keyword information.

3. The method of claim 1, wherein, the step of obtaining a user audio signal comprises: obtaining an audio signal; performing noise reduction processing on the audio signal to obtain a noise-reduced audio signal; based on a pre-set adaptive filtering algorithm, performing filtering processing on an echo signal in the noise-reduced audio signal to obtain the user audio signal.

4. The method of claim 3, wherein, The speech recognition model comprises a first recognition model and a second recognition model, and the step of inputting the user audio signal into a speech recognition model to obtain audio text information comprises: inputting the user audio signal into the first recognition model to obtain pinyin text information; wherein the first recognition model is trained by using a plurality of pre-collected audios as training inputs and audio pinyins obtained by performing pinyin conversion on the plurality of audios as target outputs; inputting the pinyin text information into the second recognition model to obtain the audio text information.

5. The method of claim 4, wherein, The first recognition model comprises a convolution module, a recurrent neural network module and a fully connected module, and the step of inputting the user audio signal into the first recognition model to obtain pinyin text information comprises: inputting the user audio signal into the first recognition model, extracting audio features in the user audio signal through the convolution module, and obtaining encoding features corresponding to the audio features; performing recurrent convolution sequence processing on the encoding features through the recurrent neural network module to obtain time sequence features; integrating the time sequence features through the fully connected module to obtain the pinyin text information.

6. A voice control method, characterized by, The method comprises the following steps: Receiving an audio signal collection trigger instruction; In response to the audio signal collection trigger instruction, an audio signal is acquired; According to the audio signal, the processed intent category information and the processed keyword information are acquired; Based on the processed intent category information and the processed keyword information, the target system is controlled; According to the audio signal, the processed intent category information and the processed keyword information are acquired, including: the audio signal is sent to the server, so that the server returns the processed intent category information and the processed keyword information, wherein the server is configured to acquire user audio according to the audio signal, perform voice intent recognition on the user audio signal to obtain intent category information and keyword information, and perform parameterized mapping processing on the intent category information and the keyword information to obtain the processed intent category information and the processed keyword information; The voice intent recognition on the user audio signal includes: The user audio signal is input into a speech recognition model to obtain audio text information; Based on a preset text mapping table, the valid text information corresponding to the audio text information is acquired; The valid text information is input into an intent recognition model to obtain the intent category information and the keyword information. The intent category information includes intent information for BIM system control; the keyword information includes word slot keywords set for each intent category.

7. The method of claim 6, wherein, The step of acquiring the audio signal in response to the audio signal collection trigger instruction includes: In response to the audio signal collection operation instruction; If the audio signal collection operation instruction is a voice wake-up instruction, the wake-up word is acquired, and after detecting that the wake-up word matches the preset valid wake-up word, the corresponding audio signal is acquired; If the audio signal collection operation instruction is a key input instruction, after detecting that the preset voice input key is triggered, the corresponding audio signal is acquired; If the audio signal collection operation instruction is a voice recognition instruction, the audio frame to be recognized is acquired, and the audio frame not containing a silent frame is selected as the corresponding audio signal.

8. The method of claim 6, wherein, The step of controlling the target system based on the processed intent category information and the processed keyword information includes: Based on the processed intent category information, a target control function is determined; The target control function is called, and the processed keyword information is input as information of the target control function to control the target system.

9. A voice control device, characterized in that It includes: A signal acquisition module is configured to acquire a user audio signal; An intent recognition module is configured to perform voice intent recognition on the user audio signal to obtain intent category information and keyword information; A parameter processing module is configured to perform parameterized mapping processing on the intent category information and the keyword information to obtain processed intent category information and processed keyword information, which are used to control a target system. The voice intent recognition on the user audio signal to obtain the intent category information and the keyword information comprises: inputting the user audio signal into a voice recognition model to obtain audio text information; obtaining valid text information corresponding to the audio text information based on a preset text mapping table; inputting the valid text information into an intent recognition model to obtain the intent category information and the keyword information; The intent category information comprises intent information for BIM system control; and the keyword information comprises slot keywords set for each intent category.

10. A voice control device, characterized in that, Comprise: An instruction receiving module configured to receive an audio signal collection trigger instruction; An instruction responding module configured to obtain an audio signal in response to the audio signal collection trigger instruction; A signal sending module configured to obtain processed intent category information and processed keyword information based on the audio signal; A system control module configured to control a target system based on the processed intent category information and the processed keyword information; The obtaining of the processed intent category information and the processed keyword information based on the audio signal comprises: sending the audio signal to a server to enable the server to return the processed intent category information and the processed keyword information, wherein the server obtains user audio based on an audio signal, obtains intent category information and keyword information by performing voice intent recognition on the user audio signal, and performs parameterized mapping processing on the intent category information and the keyword information to obtain the processed intent category information and the processed keyword information. The voice intent recognition on the user audio signal to obtain the intent category information and the keyword information comprises: inputting the user audio signal into a voice recognition model to obtain audio text information; obtaining valid text information corresponding to the audio text information based on a preset text mapping table; inputting the valid text information into an intent recognition model to obtain the intent category information and the keyword information; The intent category information comprises intent information for BIM system control; and the keyword information comprises slot keywords set for each intent category.

11. A computer device, characterized by Comprise: One or more processors; Memory; And one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the processor to implement the voice control method in any one of claims 1 to 8.

12. A computer-readable storage medium, characterized in that, A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps in the voice control method in any one of claims 1 to 8.

Images