Voice control tuning method and system based on AI knowledge base and debugging system

Abstract

The application discloses a voice control tuning method and system based on an AI knowledge base and a debugging system. The voice control tuning method comprises the following steps: collecting a natural language tuning instruction of a user through a debugging system client, and recognizing the natural language tuning instruction through a voice recognition engine to obtain recognized text; sending control information containing the recognized text to an AI knowledge base server by the debugging system client; generating a structured control scheme according to the control information by the AI knowledge base server; forming instruction information according to the structured control scheme, and sending the instruction information to an audio processor for execution through a byte stream protocol by the debugging system client; and forming an execution result according to the execution feedback result of the audio processor, and returning the execution result to the AI knowledge base server by the debugging system client. The application realizes voice control tuning of the audio processor.

Description

Technical Field

[0001] This invention relates to the field of audio processing and human-computer interaction technology, and in particular to a voice control tuning method and system based on an AI knowledge base and a debugging system. Background Technology

[0002] Currently, audio processors are configured by technicians or end users using a configuration system client. However, existing configuration system clients contain multiple interfaces and a large number of adjustable parameters. For users unfamiliar with audio processors, it is often difficult to quickly locate the interface and parameters that need adjustment. This presents a high barrier to entry for untrained users, and the operation requires both hands, making it difficult to synchronize with live performances, teaching sessions, and other scenarios. Some sound mixing systems use fixed-grammar voice command templates, which have poor compatibility with colloquial speech, misspellings, and punctuation-segmented speech. Users must strictly adhere to preset sentence structures and cannot answer query-based questions. Some systems use general-purpose language models for voice assistants, but these models typically run on web pages, do not connect to professional audio processors, do not issue device-level controls, and require an internet connection, making them unusable in weak or no-network environments. Furthermore, existing speech recognition systems are mostly trained on general corpora, resulting in low recognition rates for industry-specific terms such as device models and parameter names. Their hot word function also relies on an internet connection and is ineffective in offline mode. Furthermore, existing large models are designed to understand multiple outputs of free text, making it difficult to directly drive device control. Even with structured outputs, they typically lack engineering-level constraints such as the separation of confirmed and unconfirmed traffic, risk classification, full-link auditing of actions, and isolation of multiple client instances.

[0003] Therefore, there is an urgent need to address issues such as the high barrier to entry for tuning existing audio processors, the lack of industry terminology in general voice assistants, the requirement for internet connectivity, and the inability of intent output to directly drive devices. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings and defects of existing audio tuning technologies and provide a voice-controlled audio tuning method and system based on an AI knowledge base and tuning system. Without modifying hardware or introducing new physical control components, it utilizes a proprietary closed-source online / offline speech recognition module, a unified intent entry point on the AI ​​knowledge base server, and a structured control scheme to allow users to directly tune the audio processor using natural language. This enables non-engineers to achieve a voice-controlled audio tuning experience, providing multi-dimensional audio tuning capabilities such as voice volume control, voice switching of scene modes, and voice querying of current parameters.

[0005] According to one aspect of this application, a voice control tuning method based on an AI knowledge base and a debugging system is provided, which is implemented through interaction between an AI knowledge base server and a debugging system client, including the following steps: The user's natural language tone control commands are collected through the debugging system client and recognized by the speech recognition engine to obtain the recognized text; The debugging system client will send control information, including at least the recognized text, to the AI ​​knowledge base server; The AI ​​knowledge base server generates a structured control scheme based on the control information; The debugging system client maps instruction information according to the structured control scheme and the local protocol mapping table, and sends it to the audio processor for execution via the byte stream protocol; The debugging system client generates an execution feedback result based on the execution result returned by the audio processor and returns it to the AI ​​knowledge base server.

[0006] The speech recognition engine is a closed-source speech recognition engine that adopts an end-to-end neural network structure and is obtained by performing secondary training on the target speech recognition model based on a preset corpus. The preset corpus includes at least the audio processor's instruction manual, debugging manual, customer Q&A information about the audio processor, and transcripts of on-site recordings related to audio processor tuning; the speech recognition engine's decoder is configured with preset information related to the audio processor, which includes at least the audio processor model, adjustable parameter names, technical terms related to audio processor tuning settings, and corresponding colloquial aliases.

[0007] The speech recognition engine includes an offline speech recognition engine and an online speech recognition engine, which are selected and invoked by the user in the AI ​​assistant interface of the debugging system client.

[0008] The online speech recognition engine is deployed on a private cloud and provides recognition services to the outside world through a controlled encrypted link; the offline speech recognition engine is delivered as a local dynamic library and integrated into the debugging system client.

[0009] The debugging system client has at least one microphone trigger entry and a voice acquisition module selection entry set in its AI assistant interface.

[0010] The control information includes a client identifier and device context information. The client identifier is randomly generated by the debugging system client and is used by the AI ​​knowledge base server to isolate sessions, messages and action records, so that multiple debugging system clients can use the same AI knowledge base server concurrently without crosstalk.

[0011] The step of generating the structured control scheme includes: The identified text and device capability aliases are normalized; based on the normalization structure, the identified text is matched with device capability aliases; based on the longest normalized device capability alias among the multiple obtained device capability aliases, the executable device capability is matched; based on the matched executable device capability, the structured control scheme is output. Preferably, if no executable device capability is matched, knowledge retrieval and large model answering are performed and natural language question answering results are returned to the debugging system client; Preferably, after matching the executable device capability, the action value is extracted from the numerical value in the identified text according to the corresponding device capability value type, and the extracted action value is cropped according to the minimum and maximum values ​​of the parameters to determine the target action value. Preferably, the action value extraction includes at least one of Boolean extraction, numerical extraction, and enumeration extraction; Preferably, the numerical values ​​include at least Chinese numerals, Arabic numerals, and natural language expressions of negative numbers; Preferably, the normalization process includes at least one operation: removing whitespace characters, standardizing the case of numerical values, and removing full-width and half-width punctuation and pause marks.

[0012] The structured control scheme includes at least an action identifier, whether confirmation is pending, risk level, device model, and at least one action; the at least one action includes at least a parameter key, parameter name, action type, action value, and scope; the action type includes at least a setting action and a query action; the action identifier is stored in the session storage and action storage of the AI ​​knowledge base server and the execution log of the debugging system client, and is used to trace the complete execution chain from the recognized text to the device readback result.

[0013] Preferably, the setting action is configured to require confirmation by default, and is sent to the audio processor by the debugging system client for execution after user confirmation; the query action is configured to require no confirmation by default, and is sent directly to the audio processor by the debugging system client for execution. Preferably, the debugging system client displays the setting actions and their corresponding risk levels in a visual manner on its AI assistant interface, and each setting action is equipped with a confirmation button for user confirmation.

[0014] According to a second aspect of this application, a voice control tuning system based on an AI knowledge base and a debugging system is provided, including an AI knowledge base server deployed on a server and a debugging system client running on a computer. The computer is communicatively connected to an audio processor and the server. The voice recognition engine includes an offline voice recognition engine deployed on the debugging system client and an online voice recognition engine deployed on a private cloud. The online recognition engine is communicatively connected to the debugging system client. The debugging system client is used to collect the user's natural language tuning commands, send control information containing recognized text to the AI ​​knowledge base server; and map command information according to the structured control scheme output by the AI ​​knowledge base server and the local protocol mapping table, send it to the audio processor for execution via byte stream protocol, and form an execution result based on the feedback result returned by the audio processor, and return it to the AI ​​knowledge base server. The offline or online speech recognition engine is used to recognize the natural speech command offline or online to obtain the recognized text; The AI ​​knowledge base server is used to generate the structured control scheme based on the control information and return it to the debugging system client.

[0015] The voice control tuning technology based on an AI knowledge base and debugging system of the present invention collects the user's natural speech commands through the debugging system client. After speech recognition by the speech recognition engine, the control information, including the recognized text, is sent to the AI ​​knowledge base server for processing. The AI ​​knowledge base server returns a structured control scheme. The debugging system client maps the structured scheme to form instruction information according to the local protocol mapping table, sends it to the audio processor for execution via the byte stream protocol, and obtains the execution status and readback value. The execution result is then returned to the AI ​​knowledge base server, realizing voice control tuning. This allows users to directly tune the audio processor using natural language, enabling non-engineers to experience voice control tuning. Attached Figure Description

[0016] Figure 1 This is an architecture diagram of the voice control tuning system based on an AI knowledge base and debugging system of the present invention.

[0017] Figure 2 This is a flowchart of the voice control tuning method based on an AI knowledge base and debugging system of the present invention. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0019] See Figure 1 , Figure 2 As shown in the exemplary embodiment of this application, a voice control tuning method based on an AI knowledge base and a debugging system is provided. This method is implemented through interaction between an AI knowledge base server and a debugging system client, and includes the following steps: S1. The user's natural language tone control commands are collected through the debugging system client and recognized by the speech recognition engine to obtain the recognized text; S2. The debugging system client sends control information containing the recognized text to the AI ​​knowledge base server; S3. The AI ​​knowledge base server generates a structured control scheme based on the control information; S4. The debugging system client maps instruction information according to the structured control scheme and the local protocol mapping table, and sends it to the audio processor for execution via the byte stream protocol; S5. The debugging system client generates an execution result based on the feedback result returned by the audio processor and returns it to the AI ​​knowledge base server.

[0020] According to an embodiment of this application, a computer running the debugging system client is connected to the audio processor via a serial port or network port, and communicates with the server of the AI ​​knowledge base server via a remote interface.

[0021] In one embodiment, the structured control scheme includes at least one action, which includes at least a parameter key. The action can be a setting action or a query action. The local protocol mapping table includes the mapping relationship between the action and the parameter key. After obtaining the structured control scheme, the hardware control instruction is determined by the parameter key according to the local protocol mapping table. Combined with other action parameter information (such as action type, action value, and scope), instruction information is formed and sent to the audio processor for execution via serial port / network port byte stream protocol.

[0022] In this embodiment, the recognized text obtained through speech recognition is sent to the AI ​​knowledge base server, which determines a feasible control scheme based on the recognized text, forming a structured control scheme. Then, the debugging system client maps and converts the information in the structured control scheme into instruction information according to the local protocol mapping table, and sends it to the audio processor for execution via serial port or network port byte stream protocol. The execution result of the audio processor is then sent back to the AI ​​knowledge base server, thus forming a complete closed loop from voice instruction to execution result feedback.

[0023] According to an embodiment of this application, the speech recognition engine is a closed-source speech recognition engine, which adopts an end-to-end neural network structure and is obtained by performing secondary training on the target speech recognition model based on a preset corpus.

[0024] The preset corpus includes at least the audio processor's instruction manual, the audio processor's debugging manual, customer Q&A information about the audio processor, and on-site recordings transcribed related to the audio processor's tuning. The on-site recordings transcribed include recordings of actual sound engineers speaking their operational scripts, debugging dialogues, and equipment communication voices at audio tuning sites such as conference rooms, stages, classrooms, and performances. The audio is then converted into plain text using a speech recognition tool, and the resulting text material is the on-site recording transcribed text.

[0025] In this application, by selecting corpus information related to the audio processor itself for secondary training of the speech recognition engine, the speech recognition engine becomes more focused on speech recognition within the audio processor domain, improving the accuracy and professionalism of speech recognition. Due to the use of industry-specific corpus for secondary training, the speech recognition engine or module of this invention significantly reduces the error rate in professional expressions related to audio processor tuning compared to general speech recognition engines or modules, resulting in an improved recognition rate.

[0026] Furthermore, in this application, the decoder of the speech recognition engine is configured with preset information related to the audio processor. This preset information includes at least the audio processor model, adjustable parameter names, technical terms related to the audio processor's tuning settings, and corresponding colloquial aliases. In some embodiments, the technical terms related to the audio processor's tuning settings and corresponding colloquial aliases at least include technical terms related to adjustable parameters, device capabilities, and tuning intentions, along with their colloquial aliases. By setting this preset information, the speech recognition capability of the speech recognition engine can be effectively improved.

[0027] According to embodiments of this application, the speech recognition engine (such as an ASR automatic speech recognition model) includes an offline speech recognition engine and an online speech recognition engine, which are selected and invoked by the user through the AI ​​assistant interface of the debugging system client, allowing the user to conveniently choose between the offline and online speech recognition engines. The offline engine supports the acquisition of recognized text when the computer is completely offline, and enables the debugging system client to still complete text recognition and all steps up to the distribution of the structured control scheme even when offline.

[0028] According to an embodiment of this application, the online speech recognition engine can be deployed on a private cloud, providing recognition services to the outside world through a controlled encrypted link, interacting with the debugging system client, receiving speech information from the debugging system client, recognizing it, and returning it to the debugging system client; the offline speech recognition engine can be delivered as a local dynamic library and integrated into the debugging system client.

[0029] In this embodiment, the debugging system client includes a voice acquisition module, a voice recognition engine adaptation layer, an AI assistant interface, a knowledge base remote communication module, a device protocol mapping module, and a serial port / network port transmission module. The voice acquisition module is used to acquire natural language signals used by the user for audio tuning. The AI ​​assistant interface provides an interactive interface for the user to input information and output preset information. The knowledge base remote communication module is used to communicate locally or remotely with the AI ​​knowledge base server. The serial port / network port transmission module is used to communicate with the audio processor. The device protocol mapping module is used to map and convert information in the structured control scheme into instruction information according to the local protocol mapping table, and send it to the audio processor for execution via serial port or network port byte stream protocol. The voice recognition engine adaptation layer is used for mode selection and switching of the voice recognition engine.

[0030] According to an embodiment of this application, the debugging system client has at least one microphone trigger entry and one voice acquisition module selection entry set in its AI assistant interface; the microphone trigger entry is a non-hardware virtual trigger entry, such as displaying a virtual sound input icon in the AI ​​assistant interface, which is touched by the user to trigger microphone acquisition.

[0031] In some embodiments, the AI ​​assistant interface of the debugging system client may have a dialogue display area in the middle to display human-computer dialogue records, and the left or right side may display historical records, forming an intuitive dialogue display, which makes it convenient for users to grasp the current conversation and historical conversation records. The conversation or dialogue is stored and recorded by the AI ​​knowledge base server in the backend.

[0032] In some embodiments, a text input sending entry is also provided in the AI ​​assistant interface. The text input sending entry may include a text input box and a text sending button. The text sending button is a non-hardware virtual trigger button displayed by the debugging system client program on its AI assistant interface. After the user touches the trigger button, the text content of the input text box can be sent. The user can input relevant request information through text, which may include tuning command type information or question and answer type information.

[0033] According to embodiments of this application, the voice acquisition module includes a built-in / external voice acquisition module of the computer, including at least one of a built-in / external sound card, a voice acquisition module presented to the computer's operating system after the audio processor is connected to the computer via a serial port or network port, and so on. By using the voice acquisition module presented through the audio processor, audio can be supplied to the debugging system client through the multi-channel microphone of the audio processor. Specifically, a drop-down menu for voice acquisition modules can be provided on the AI ​​assistant interface to allow selection of different voice acquisition modules.

[0034] According to an embodiment of this application, the control information includes a client identifier; the client identifier is randomly generated by the debugging system client upon initial startup and is used by the AI ​​knowledge base server to isolate sessions, messages, and action records, enabling multiple debugging system clients to concurrently use the same AI knowledge base server without crosstalk.

[0035] In some embodiments, the control information also includes query-type information, namely device context information. The device context information includes, but is not limited to, the real-time parameters of the audio processor, audio processor model, firmware version, set of supported device capability aliases (channel / parameter adjustable), protocol type, number of channels, currently selected routing matrix, and current parameter values ​​such as gain, delay, and compression threshold for each channel and frequency band; whether feedback suppression, frequency division, high-pass and low-pass filtering, and other device function activation status are enabled; and device hardware status such as channel occupancy, phase mode, and output switch status.

[0036] According to embodiments of this application, after receiving the identified text, the AI ​​knowledge base server matches and judges the information in the identified text with the alias set in the knowledge base. Based on the matched target alias, it determines the feedback result to be returned to the debugging system client. The feedback result includes a structured control scheme. In some embodiments, if the user-input voice information is not a tuning command of the audio processor, a fixed answer is returned after matching a fixed answer. If it is a tuning command, a structured control scheme is output after matching with device capability aliases. If no fixed answer or control scheme is matched, natural language knowledge question-and-answer information is returned.

[0037] In some embodiments, the alias is a device capability alias; the AI ​​knowledge base server matches the corresponding alias according to the recognized text, and determines the corresponding feedback result to the debugging system client based on the matched alias. There can be multiple device capability aliases, which may include standard tuning terms, tuning terms formed in different languages, different colloquialisms, or local languages, etc. Multiple device capability aliases correspond to a professional term in the tuning field. This enables the matching of corresponding tuning professional terms based on different speech recognition results, so that the AI ​​knowledge base server can accurately match the corresponding tuning professional terms by parsing the content of the recognized text.

[0038] In this application, the AI ​​knowledge base server may be configured with a fixed answer module, a control planner, a knowledge retrieval and answering module, and a conversation and action storage module to perform the above functions respectively. For example, the fixed answer module outputs a fixed answer that matches the recognized text based on the recognized text; the control planner outputs a structured control scheme based on the recognized text; the knowledge retrieval and answering module performs knowledge retrieval based on the recognized text, and uses a large model to answer based on the knowledge retrieval results, returning natural language question-and-answer results; the conversation and action storage module is used to store the conversation and action information generated during the tuning process and archive it.

[0039] Furthermore, in some embodiments, the AI ​​knowledge base server can be configured to, upon receiving the recognized text from the speech recognition engine, first perform normalization processing on both the recognized text and the device capability alias. Based on the normalization processing result, the normalized device capability alias is scanned and matched with the recognized text to determine the intent of the recognized text. Through the normalization processing operation, punctuation and whitespace are normalized to tolerate recognition fragments, ensuring that recognition results with sentence-ending punctuation, sentence pauses, or whitespace (e.g., "Cancel. Mute." with a period) can still match aliases without punctuation (e.g., "Cancel Mute"), thus accurately recognizing the intent to adjust the tone. According to embodiments of this application, the normalization processing includes removing whitespace characters, standardizing the case of numerical values, and removing full-width and half-width punctuation such as commas, periods, pauses, semicolons, colons, exclamation marks, question marks, parentheses, and quotation marks.

[0040] Furthermore, after identifying the intent information of the identified text, the system first scans and matches a preset fixed answer based on the intent information. If the match is successful, the preset fixed answer is returned to the debugging system client to answer the user's question. The user's knowledge acquisition needs are met through the preset fixed answer.

[0041] In some embodiments, if a preset fixed answer cannot be matched based on the intent information of the identified text, a matching scan is performed between the normalized identified text and the device capability alias. Based on the longest normalized device capability alias among the multiple successfully matched device capability aliases, the executable device capability (i.e. the device's executable control capability, such as adjustable functions / parameters including volume adjustment, scene mode switching, equalizer parameter setting, mute, etc.) is scanned for and matched. If found, the structured control scheme is output based on the executable device capability and the device context information.

[0042] Furthermore, in this application, if no device capability is available, knowledge retrieval and large model answering are performed and natural language question-answering results are returned to the debugging system client for user use. When the debugging software client receives the natural language question-answering results, it only displays and saves them without performing any device control.

[0043] According to an embodiment of this application, after scanning the executable device capability, further, based on the value type of the corresponding device capability, action values ​​are extracted from the numerical values ​​(including Chinese numerals, Arabic numerals, and natural language expressions of negative numbers) in the identified text. The extracted action values ​​are then trimmed according to the minimum and maximum values ​​of the parameters to determine the target action value, ultimately forming a structured control scheme. The action value extraction includes at least one of Boolean extraction, numerical extraction, and enumeration extraction.

[0044] The debugging system client of this application uses a unified entry point for knowledge-based Q&A and control parameters related to the audio processor. The same interface handles both knowledge-based Q&A and parameter control requests. When a structured control scheme cannot be output for execution, it seamlessly returns to knowledge retrieval and large model Q&A, providing users with a unified experience of "one input box, two types of capabilities".

[0045] In this application, the device capabilities include adjustable functional parameters of the audio processor, such as input channel capabilities, output channel capabilities, system-level capabilities, and matrix / routing capabilities.

[0046] In addition, this application combines the normalization of the identified text and device capability aliases with the longest device capability alias matching, which has a strong tolerance for identification fragments and colloquialisms, and achieves highly robust colloquialism tolerance.

[0047] According to embodiments of this application, the structured control scheme includes at least an action identifier, whether confirmation is pending, risk level, device model, and at least one action; the at least one action includes at least a parameter key, parameter name, action type, action value, and scope; the action type includes at least setting actions and query actions, and there are multiple actions, which can be displayed in a list format. In this application, the action identifier is a globally unique identifier that runs through the session storage and action storage of the AI ​​knowledge base server and the execution log of the debugging system client, used to trace the complete execution chain from the recognized text to the device readback result.

[0048] In this embodiment of the application, the setting action is configured by default to pending confirmation, and after user confirmation, it is sent to the audio processor by the debugging system client for execution; the query action is configured by default to no confirmation required, and is sent directly to the audio processor by the debugging system client for execution.

[0049] According to embodiments of this application, the debugging system client displays the setting actions and their corresponding risk levels in a visual manner on its AI assistant interface. Each setting action is configured with a virtual confirmation button for user confirmation, allowing the user to confirm whether to continue based on the risk level. It can also be set or configured for the debugging system client to provide secondary confirmation prompts via pop-up windows. In some embodiments, the risk levels include at least three levels: low, medium, and high. When the control scheme includes multiple actions, the risk level is the most severe level that is displayed.

[0050] According to embodiments of this application, the feedback result includes at least the execution status information and readback value returned by the response setting action, and the current parameter information returned by the response query action; the execution result includes at least the feedback result, and may also include the original byte response data corresponding to the feedback result, communication protocol name, and other information. The execution result can be returned to the AI ​​knowledge base client through a preset action result feedback interface, and the AI ​​knowledge base client archives the session and action status according to the client identifier.

[0051] In this application, the AI ​​knowledge base server can be deployed locally or in the cloud. If it is deployed locally to form a local AI knowledge base server, the technology of this application, through the combination of an offline speech recognition engine, a local AI knowledge base server and a local protocol mapping table, can also complete the entire process of voice control tuning of the audio processor even when the whole machine is offline.

[0052] In this application, the AI ​​knowledge base server and the debugging system client are composed of computer programs, stored in a memory, and called by the processor. They interact with each other in conjunction with the speech recognition engine, thereby realizing the voice control and tuning of the audio processor.

[0053] According to a second aspect of this application, a voice control tuning system based on an AI knowledge base and a debugging system is provided, including an AI knowledge base server deployed on a server and a debugging system client running on a computer, wherein the computer is communicatively connected to an audio processor and the server; wherein the voice recognition engine includes an offline voice recognition engine deployed on the debugging system client and an online voice recognition engine deployed on a private cloud; the online recognition engine is communicatively connected to the debugging system client; The debugging system client is used to collect the user's natural language tuning commands, send control information containing recognized text to the AI ​​knowledge base server; and map command information according to the structured control scheme output by the AI ​​knowledge base server and the local protocol mapping table, send it to the audio processor for execution via byte stream protocol, and form an execution result based on the feedback result returned by the audio processor, and return it to the AI ​​knowledge base server. The offline or online speech recognition engine is used to recognize the natural speech command offline or online to obtain the recognized text; The AI ​​knowledge base server is used to generate the structured control scheme based on the control information and return it to the debugging system client.

[0054] The voice-controlled tuning technology of this application, through the application of AI knowledge base server and debugging system client, can lower the tuning threshold of audio processor. Users can control the tuning of audio processor through natural language. Even untrained on-site users can directly complete the professional parameter adjustment of audio processor. It is also compatible with existing debugging systems or programs: it can be implemented as a new module in the original debugging system client or system without affecting the original manual parameter adjustment process. The audio processor can be adapted without modification.

[0055] The technology in this application ensures security, observability, and multi-terminal availability in the production environment by setting up a combination of confirmation-required and confirmation-free traffic separation, risk classification, action identifier auditing, and client identifier isolation in the help interface.

[0056] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, the embodiments should be regarded as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, it is intended that all variations falling within the meaning and scope of the equivalents of the claims be included within the invention.

[0057] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A voice control tuning method based on an AI knowledge base and debugging system, characterized in that, This implementation, based on the interaction between the AI ​​knowledge base server and the debugging system client, includes the following steps: The user's natural language tone control commands are collected through the debugging system client and recognized by the speech recognition engine to obtain the recognized text; The debugging system client sends control information containing the recognized text to the AI ​​knowledge base server; The AI ​​knowledge base server generates a structured control scheme based on the control information; The debugging system client maps instruction information according to the structured control scheme and the local protocol mapping table, and sends it to the audio processor for execution via the byte stream protocol; The debugging system client generates an execution result based on the feedback returned by the audio processor and returns it to the AI ​​knowledge base server.

2. The voice control tuning method based on an AI knowledge base and debugging system according to claim 1, characterized in that, The speech recognition engine is a closed-source speech recognition engine, which adopts an end-to-end neural network structure and is obtained by performing secondary training on the target speech recognition model based on a preset corpus. The preset corpus includes at least the audio processor's instruction manual, debugging manual, customer Q&A information about the audio processor, and transcripts of on-site recordings related to audio processor tuning; The decoder of the speech recognition engine is configured with preset information related to the audio processor. The preset information includes at least the audio processor model, adjustable parameter names, technical terms related to the audio processor tuning settings, and corresponding colloquial aliases.

3. The voice control tuning method based on an AI knowledge base and debugging system according to claim 1, characterized in that, The speech recognition engine includes an offline speech recognition engine and an online speech recognition engine, which can be selected and invoked by the user in the AI ​​assistant interface of the debugging system client.

4. The voice control tuning method based on an AI knowledge base and debugging system according to claim 3, characterized in that, The online speech recognition engine is deployed on a private cloud and provides recognition services to the outside world through a controlled encrypted link; the offline speech recognition engine is delivered as a local dynamic library and integrated into the debugging system client.

5. The voice control tuning method based on an AI knowledge base and debugging system according to claim 1, characterized in that, The debugging system client has at least one microphone trigger entry and a voice acquisition module selection entry set in its AI assistant interface.

6. The voice control tuning method based on an AI knowledge base and debugging system according to claim 1, characterized in that, The control information includes the client identifier and device context information; The client identifier is randomly generated by the debugging system client and is used by the AI ​​knowledge base server to isolate sessions, messages and action records, so that multiple debugging system clients can use the same AI knowledge base server concurrently without crosstalk.

7. The voice control tuning method based on an AI knowledge base and debugging system according to claim 1, characterized in that, The steps for generating the structured control scheme include: The identified text and device capability aliases are normalized; based on the normalization structure, the identified text is matched with device capability aliases; based on the longest normalized device capability alias among the multiple obtained device capability aliases, the executable device capability is matched; based on the matched executable device capability, the structured control scheme is output. Preferably, if no executable device capability is matched, knowledge retrieval and large model answering are performed and natural language question answering results are returned to the debugging system client; Preferably, after matching the executable device capability, the action value is extracted from the numerical value in the identified text according to the corresponding device capability value type, and the extracted action value is cropped according to the minimum and maximum values ​​of the parameters to determine the target action value. Preferably, the action value extraction includes at least one of Boolean extraction, numerical extraction, and enumeration extraction; Preferably, the numerical values ​​include at least Chinese numerals, Arabic numerals, and natural language expressions of negative numbers; Preferably, the normalization process includes at least one operation: removing whitespace characters, standardizing the case of numerical values, and removing full-width and half-width punctuation and pause marks.

8. The voice control tuning method based on an AI knowledge base and debugging system according to claim 1, characterized in that, The structured control scheme includes at least an action identifier, whether confirmation is pending, risk level, equipment model, and at least one action; at least one action includes at least a parameter key, parameter name, action type, action value, and scope; the action type includes at least a setting action and a query action; The action identifier is stored throughout the session storage and action storage of the AI ​​knowledge base server and the execution log of the debugging system client, and is used to trace the complete execution chain from the recognized text to the device readback result.

9. The voice control tuning method based on an AI knowledge base and debugging system according to claim 8, characterized in that, The default setting for the action is to await confirmation. After user confirmation, the action is sent from the debugging system client to the audio processor for execution. The query-type action is configured to be executed without confirmation by default, and is directly sent from the debugging system client to the audio processor for execution. Preferably, the debugging system client displays the setting actions and their corresponding risk levels in a visual manner on its AI assistant interface, and each setting action is equipped with a confirmation button for user confirmation.

10. A voice control tuning system based on an AI knowledge base and debugging system, characterized in that, It includes an AI knowledge base server deployed on a server, a debugging system client running on a computer, and a speech recognition engine. The computer is communicatively connected to an audio processor and the server. The speech recognition engine includes an offline speech recognition engine deployed on the debugging system client and an online speech recognition engine deployed on a private cloud. The online recognition engine communicates with the debugging system client. The debugging system client is used to collect the user's natural language tuning commands, send control information containing recognized text to the AI ​​knowledge base server; and map command information according to the structured control scheme output by the AI ​​knowledge base server and the local protocol mapping table, send it to the audio processor for execution via byte stream protocol, and form an execution result based on the feedback result returned by the audio processor, and return it to the AI ​​knowledge base server. The offline or online speech recognition engine is used to recognize the natural speech command offline or online to obtain the recognized text; The AI ​​knowledge base server is used to generate the structured control scheme based on the control information and return it to the debugging system client.

Images