Intelligent Bluetooth voice remote control system based on AI semantic analysis

Abstract

The invention relates to the technical field of intelligent voice interaction, in particular to an intelligent Bluetooth voice remote control system based on AI semantic analysis. Comprising a voice acquisition unit, a Bluetooth communication unit, a voice recognition unit, an AI semantic analysis unit and a control execution unit, and the Bluetooth communication unit is used for establishing low-power-consumption Bluetooth connection with target equipment and supporting bidirectional data transmission. By combining advanced localized model library, edge calculation optimization, multi-modal data fusion and dynamic semantic map technologies, the recognition accuracy, the real-time response speed and the context understanding ability of the voice instruction are remarkably improved, and meanwhile, the data privacy is guaranteed, so that efficient, personalized, natural and smooth user interaction experience is realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent voice interaction technology, and specifically to an intelligent Bluetooth voice remote control system based on AI semantic analysis. Background Art

[0002] Smart Bluetooth Voice is an interactive system that combines advanced artificial intelligence (AI) technology with Bluetooth wireless communication to provide users with a convenient and efficient voice control experience. By integrating high-precision automatic speech recognition, natural language processing, and semantic understanding algorithms, Smart Bluetooth Voice accurately converts user voice commands into executable commands and transmits them to target devices stably and quickly via Bluetooth low energy technology. This system not only supports multilingual and dialect adaptation but also provides personalized responses and services based on user habits and preferences. Furthermore, with edge computing optimization and a localized model library, Smart Bluetooth Voice significantly reduces latency and improves real-time performance while ensuring data privacy. Its application scenarios are wide-ranging, including but not limited to smart home control, in-vehicle infotainment systems, and remote operation of various smart devices. With Smart Bluetooth Voice, users can easily and seamlessly control multiple devices, enjoying a more intelligent and convenient lifestyle. This technological advancement not only enhances the naturalness and efficiency of human-computer interaction but also provides strong support for the interconnection and interoperability of various smart devices. In the existing technology, traditional voice remote control systems have low accuracy in recognizing dialects and professional terms. When processing complex voice commands, it often takes a long time to complete voice recognition and semantic analysis. Especially in poor network conditions, the user experience is seriously affected, and there are also problems such as insufficient context understanding ability and data privacy risks.

[0003] Based on this, the present invention provides an intelligent Bluetooth voice remote control system based on AI semantic analysis to solve the above-mentioned technical problems. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent Bluetooth voice remote control system based on AI semantic analysis. By combining advanced localization model libraries, edge computing optimization, multimodal data fusion and dynamic semantic graph technology, the present invention significantly improves the recognition accuracy, real-time response speed and context understanding ability of voice commands, while ensuring data privacy, thereby achieving an efficient, personalized and natural user interaction experience.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] The present invention provides an intelligent Bluetooth voice remote control system based on AI semantic analysis, comprising a voice acquisition unit, a Bluetooth communication unit, a voice recognition unit, an AI semantic analysis unit, and a control execution unit, wherein:

[0007] The voice collection unit is used to collect the user's voice instructions and perform pre-processing;

[0008] The Bluetooth communication unit is used to establish a low-power Bluetooth connection with the target device and support two-way data transmission;

[0009] The speech recognition unit is used to convert voice commands into text using automatic speech recognition technology, supports multi-language and dialect adaptation, and improves the recognition accuracy and real-time performance of accents and professional terms through localized model training and edge computing optimization;

[0010] The AI semantic analysis unit: Based on intelligent semantic analysis of multimodal fusion, it realizes intent understanding, ambiguity resolution and personalized interaction through dialogue state tracking, user portrait modeling and context association analysis;

[0011] The control execution unit is used to convert the parsed instructions into the control protocol supported by the target device, support Bluetooth HID, infrared coding multi-protocol adaptive conversion, and trigger execution and status feedback.

[0012] The voice acquisition unit includes a microphone array module, a preprocessing module, a voice activity detection module, and a low-power wake-up module, wherein:

[0013] The microphone array module is used to collect voice signals emitted by the user from the surrounding environment and can capture sounds from multiple directions and angles;

[0014] The pre-processing module is responsible for performing noise reduction, echo cancellation and voice enhancement on the captured voice signal;

[0015] The voice activity detection is used to identify valid voice segments and filter out silence and background noise;

[0016] The low-power wake-up module is used to wake up the device through keyword detection and reduce standby power consumption.

[0017] The Bluetooth communication unit includes a Bluetooth chip module, a radio frequency front-end module, a Bluetooth protocol stack module, and a data cache and processing module, wherein:

[0018] The Bluetooth chip module is responsible for implementing the underlying functions of the Bluetooth protocol, including signal modulation and demodulation, and frequency switching;

[0019] The RF front-end module is used to perform power amplification and filtering on Bluetooth signals to enhance the signal transmission capability and anti-interference capability;

[0020] The Bluetooth protocol stack module is used to implement various protocols and specifications of Bluetooth communication, such as pairing, connection management, and data transmission, and provides a standardized interface for upper-layer applications;

[0021] The data buffering and processing module is used to temporarily store the data to be sent and the data that has been received, and to perform necessary processing.

[0022] The speech recognition unit includes a feature extraction module, a localization model library module, an edge computing module, and a multi-language adaptation module, wherein:

[0023] The feature extraction module is used to extract representative feature parameters, such as Mel-frequency cepstral coefficients, from the preprocessed speech signal;

[0024] The localized model library module is used to store speech recognition models trained for different regions, dialects and professional fields, and select the appropriate model for recognition according to the user's usage scenario;

[0025] The edge computing module is used to perform some computing tasks on the local device, reducing data transmission delays, improving the real-time performance of recognition, and protecting user data privacy;

[0026] The multilingual adaptation module is used to support the recognition of multiple languages and dialects, and achieves accurate recognition in different language environments by switching language models and vocabularies.

[0027] The localized model library module is used to store speech recognition models trained for different regions, dialects, and professional fields. The appropriate model is selected for recognition based on the user's usage scenario. The specific operations are as follows:

[0028] A1: Automatically trigger model switching based on user location, connection device type, or time characteristics;

[0029] A2: Load pre-trained dialect / domain models, such as Cantonese and medical terminology models, from local encrypted storage. Compress the model size to ≤50MB to adapt to edge devices.

[0030] A3: When it is detected that the user has corrected the recognition result three times in a row, the output weights of the general model and the domain-specific model are automatically mixed;

[0031] A4: Use digital signatures to verify the integrity and origin of model files and prevent unauthorized model loading.

[0032] The edge computing module is used to perform some computing tasks on the local device, reduce data transmission delays, improve the real-time performance of recognition, and protect user data privacy. The specific operations are as follows:

[0033] B1: The speech recognition process is divided into acoustic feature extraction performed locally and semantic understanding processed in the cloud, where acoustic feature extraction is completed on the device side;

[0034] B2: Remove biometric information from voice features, such as voiceprint features, before uploading feature data, and retain content-related parameters;

[0035] B3: Adjust the distribution ratio of local and cloud computing tasks in real time based on network latency and computing load;

[0036] B4: Stores the acoustic model and language model of high-frequency words, and quickly retrieves them through Bloom filters, making the local recognition response time less than 200ms.

[0037] The AI semantic analysis unit includes a multimodal data fusion module, a dialogue state tracking module, a user portrait modeling module, and a context association analysis module, wherein:

[0038] The multimodal data fusion module is used to fuse data of multiple modalities such as voice, text, and image, and comprehensively analyze the user's intention and context information;

[0039] The dialogue state tracking module is used to record and track the historical information and current state of the dialogue, providing contextual support for accurately understanding user intent;

[0040] The user portrait modeling module is used to build user portraits based on the user's historical behavior, preferences and usage habits, and realize personalized semantic analysis and interaction;

[0041] The context association analysis module is used to analyze the association relationship between the current conversation and the context, eliminate semantic ambiguity, and accurately understand the user's true intention.

[0042] The user portrait modeling module is used to build user portraits based on the user's historical behavior, preferences, and usage habits, and implement personalized semantic analysis and interaction. The specific operations are as follows:

[0043] C1: Records the user's historical command sequence, device operation frequency, and time period distribution;

[0044] C2: Dynamically calculate the user's preference weight for the device based on the number of command executions and the most recent usage time;

[0045] C3: Adopts a federated learning framework to aggregate local encrypted behavior data and update user profiles weekly.

[0046] C4: When there is ambiguity in user instructions, the device or parameter with the highest weight in the profile is prioritized for execution.

[0047] The context association analysis module is used to analyze the association between the current conversation and the context, eliminate semantic ambiguity, and accurately understand the user's true intention. The specific operations are as follows:

[0048] D1: stores the command texts of the last three rounds of conversations and the environmental parameters of the device status in real time;

[0049] D2: Build a dynamic semantic graph based on the current instruction and entities in the cache context;

[0050] D3: When multiple interpretations of an instruction are detected, the optimal solution is selected based on the node association of the semantic graph;

[0051] D4: Feedback the resolution results to the voice interaction interface through confidence scoring.

[0052] The control execution unit includes a protocol conversion module, a device driver module, a status feedback module, and an error handling and retry module, wherein:

[0053] The protocol conversion module is used to convert the instructions parsed by the AI semantic analysis unit into control protocols supported by the target device, such as Bluetooth HID and infrared coding, to achieve adaptive conversion of multiple protocols;

[0054] The device driver module is used to generate a corresponding drive signal according to the converted control protocol to trigger the target device to perform a corresponding operation;

[0055] The state feedback module is used to receive the state information of the target device and feed it back to the system so that the user can understand the execution status of the device;

[0056] The error handling and retry module is used to detect and handle possible errors during the control execution process, such as communication failures and abnormal device responses, and perform retry operations.

[0057] Compared with the prior art, the present invention has the following beneficial effects:

[0058] 1. By combining advanced localized model libraries, edge computing optimization, multimodal data fusion, and dynamic semantic graph technology, this invention significantly improves the recognition accuracy, real-time response speed, and contextual understanding capabilities of voice commands while ensuring data privacy, thereby achieving an efficient, personalized, and natural user interaction experience.

[0059] 2. The present invention ensures the system's wide compatibility, operational accuracy and reliability, and improves the user experience, continuity and stability of operations by supporting adaptive conversion of multiple control protocols, generating accurate drive signals, real-time monitoring and feedback of device status, and automatic detection and repair of communication failures. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a system diagram of an intelligent Bluetooth voice remote control system based on AI semantic analysis in the present invention.

[0061] Figure 2 This is a schematic diagram of constructing a dynamic semantic graph in an intelligent Bluetooth voice remote control system based on AI semantic analysis in the present invention.

[0062] Figure 3 This is a timing diagram of multi-protocol conversion in an intelligent Bluetooth voice remote control system based on AI semantic analysis in the present invention.

[0063] Description of Figure Numbers:

[0064] 100. Voice acquisition unit; 101. Microphone array module; 102. Preprocessing module; 103. Voice activity detection; 104. Low-power wake-up module; 200. Bluetooth communication unit; 201. Bluetooth chip module; 202. RF front-end module; 203. Bluetooth protocol stack module; 204. Data cache and processing module; 300. Voice recognition unit; 301. Feature extraction module; 302. Localization model library module; 303. Edge computing module; 304. Multi-language adaptation module; 400. AI semantic analysis unit; 401. Multimodal data fusion module; 402. Dialogue state tracking module; 403. User portrait modeling module; 404. Context association analysis module; 500. Control execution unit; 501. Protocol conversion module; 502. Device driver module; 503. State feedback module; 504. Error handling and retry module. DETAILED DESCRIPTION

[0065] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0066] Example:

[0067] like Figure 1-Figure 3As shown, this embodiment provides an intelligent Bluetooth voice remote control system based on AI semantic analysis, including a voice collection unit 100, a Bluetooth communication unit 200, a voice recognition unit 300, an AI semantic analysis unit 400, and a control execution unit 500, wherein: the voice collection unit 100 is used to collect the user's voice commands and perform preprocessing; the Bluetooth communication unit 200 is used to establish a low-power Bluetooth connection with the target device and support two-way data transmission; the voice recognition unit 300 is used to convert voice commands into text through automatic speech recognition technology, support multi-language and dialect adaptation, and improve the recognition accuracy and real-time performance of accents and professional terms through localized model training and edge computing optimization; the AI semantic analysis unit 400 is based on intelligent semantic parsing of multimodal fusion, and realizes intent understanding, ambiguity resolution and personalized interaction through dialogue state tracking, user portrait modeling and context association analysis; the control execution unit 500 is used to convert the parsed commands into a control protocol supported by the target device, support multi-protocol adaptive conversion of Bluetooth HID and infrared coding, and trigger execution and status feedback.

[0068] In this embodiment, it should be noted that the AI semantic analysis unit 400 serves as the core decision-making center, bridging the text output of the voice recognition unit 300 and the instruction generation of the control execution unit 500. The Bluetooth communication unit 200 ensures low-latency data transmission throughout, forming an end-to-end intelligent interactive system.

[0069] In the present invention, the voice collection unit 100 includes a microphone array module 101, a preprocessing module 102, a voice activity detection 103, and a low-power wake-up module 104, wherein: the microphone array module 101 is used to collect voice signals emitted by the user from the surrounding environment, and can capture sounds in multiple directions and angles; the preprocessing module 102 is responsible for performing noise reduction, echo cancellation and voice enhancement on the captured voice signals; the voice activity detection 103 is used to identify valid voice segments and filter silence and background noise; the low-power wake-up module 104 is used to realize device wake-up through keyword detection and reduce standby power consumption.

[0070] In this embodiment, it should be noted that the microphone array module 101 collects the original voice signal from multiple angles, and after the preprocessing module 102 performs noise reduction and voice enhancement, the voice activity detection module 103 screens the valid voice segments, and finally the low-power wake-up module 104 realizes the device wake-up triggered by keywords.

[0071] In addition, it should be noted that the pre-processing module 102 uses spectral subtraction to reduce noise, which can maintain speech clarity in an environment with a signal-to-noise ratio of <15dB. The low-power wake-up module 104 uses a 3-layer CNN to detect keywords.

[0072] In the present invention, the Bluetooth communication unit 200 includes a Bluetooth chip module 201, a radio frequency front-end module 202, a Bluetooth protocol stack module 203, and a data cache and processing module 204, wherein: the Bluetooth chip module 201 is responsible for implementing the underlying functions of the Bluetooth protocol, including signal modulation and demodulation, and frequency switching; the radio frequency front-end module 202 is used to power amplify and filter the Bluetooth signal to enhance the signal transmission capability and anti-interference capability; the Bluetooth protocol stack module 203 is used to implement various protocols and specifications of Bluetooth communication, such as pairing, connection management, and data transmission, and provide a standardized interface for upper-layer applications; the data cache and processing module 204 is used to temporarily store data to be sent and received, and perform necessary processing.

[0073] In this embodiment, it should be noted that the Bluetooth chip module 201 implements the underlying signal processing, and after power optimization and anti-interference filtering are performed by the RF front-end module 202, the Bluetooth protocol stack module 203 completes the standardized protocol interaction, and finally the data buffering and flow control are realized through the data caching and processing module 204.

[0074] It should also be noted that the Bluetooth chip uses the Nordic nRF52840, supporting the Bluetooth 5.2 protocol with a transmission rate of 2Mbps and a communication range of up to 100 meters in open air. The RF front-end module 202 uses a power amplifier (PA) and low-noise amplifier (LNA) combination, with a PA gain of 18dB and an LNA noise figure of 2.5dB, improving signal transmission stability. The data cache and processing module 204 implements a double buffering mechanism to ensure no packet loss even with 100ms of network jitter. Essential processing within the data cache and processing module 204 includes data format conversion and error checking.

[0075] In the present invention, the speech recognition unit 300 includes a feature extraction module 301, a localization model library module 302, an edge computing module 303, and a multi-language adaptation module 304, wherein: the feature extraction module 301 is used to extract representative feature parameters, such as Mel-frequency cepstral coefficients, from the preprocessed speech signal; the localization model library module 302 is used to store speech recognition models trained for different regions, dialects, and professional fields, and select a suitable model for recognition according to the user's usage scenario; the specific operations are: A1: automatically triggering model switching through user geographic location, connection device type, or time features; A2: loading pre-trained dialect / professional field models, such as Cantonese models and medical terminology models, from local encrypted storage, and compressing the model size to ≤50MB to adapt to edge devices; A3: when it is detected that the user has corrected the recognition result three times in a row, automatically mixing the output weights of the general model and the domain-specific model; A4: verifying the integrity and source of the model file through digital signature to prevent unauthorized model loading. The edge computing module 303 is used to perform some computing tasks on the local device, reducing data transmission delays, improving the real-time performance of recognition, and protecting user data privacy. The specific operations are as follows: B1: Divide the speech recognition process into acoustic feature extraction performed locally and semantic understanding processed in the cloud, where acoustic feature extraction is completed on the device side; B2: Remove biometric information in the speech features, such as voiceprint features, before uploading the feature data, and retain content-related parameters; B3: Adjust the distribution ratio of local and cloud computing tasks in real time based on network latency and computing load; B4: Store acoustic models and language models of high-frequency words, and quickly retrieve them through Bloom filters to make the local recognition response time <200ms. The multilingual adaptation module 304 is used to support the recognition of multiple languages and dialects, and achieve accurate recognition in different language environments by switching language models and vocabularies.

[0076] In this embodiment, it should be noted that after the feature extraction module 301 converts the speech signal into feature parameters such as Mel-frequency cepstral coefficients, the localization model library module 302 dynamically selects or mixes the adapted dialect / professional model according to the user scenario. The edge computing module 303 completes the acoustic feature processing and privacy filtering locally, and collaborates with the multilingual adaptation module 304 to achieve low-latency, high-accuracy cross-language recognition.

[0077] In addition, it should be noted that by collecting voice data from various dialects such as Minnan dialect, Sichuan dialect and professional fields such as medical and industrial, the model volume is compressed using knowledge distillation technology. The specific formula is: L KD =α·L CE (y,σ(z s ))+(1-α)·L KL (σ(z s / T),(σ(z t / T)), where L CE is the cross entropy loss; T = 3 is the temperature parameter; α = 0.7 is the loss weight; z t , z s is the logits output of the teacher / student model; σ is the weight coefficient between the real label and the soft label. Dynamic switching strategy: For example, when the user is in Guangdong, the system automatically loads the Cantonese model. The dynamic allocation strategy of edge computing, the specific decision formula is: Where, α local is the allocation weight of the local computing task; e is a natural constant, approximately 2.71828; K is a hyperparameter that controls the slope of the curve; T net Network delay for real-time monitoring; E bat E is the current battery level of the edge device; max The maximum power of the edge device battery.

[0078] In the present invention, the AI semantic analysis unit 400 includes a multimodal data fusion module 401, a dialogue state tracking module 402, a user portrait modeling module 403, and a context association analysis module 404, wherein: the multimodal data fusion module 401 is used to fuse and process data of multiple modalities such as voice, text, and images, and comprehensively analyze the user's intentions and context information; the dialogue state tracking module 402 is used to record and track the historical information and current status of the dialogue, and provide context support for accurately understanding the user's intentions; the user portrait modeling module 403 is used to construct a user portrait based on the user's historical behavior, preferences, and usage habits, and realize personalized semantic analysis and interaction; the specific operations are as follows: C1: Record the user's historical command sequence, device operation frequency, and time period distribution; C2: Dynamically calculate the user's preference weight for the device based on the number of command executions and the most recent usage time; C3: Adopt a federated learning framework to aggregate local encrypted behavior data every week to update the user portrait; C4: When there is ambiguity in the user command, the device or parameter with the highest weight in the portrait is preferentially selected for execution. Contextual Relationship Analysis Module 404 analyzes the relationship between the current conversation and the context, eliminating semantic ambiguity and accurately understanding the user's true intent. The specific operations are as follows: D1: Real-time storage of command text and device status environmental parameters from the last three rounds of conversation; D2: Constructing a dynamic semantic graph based on the current command and entities in the cached context; D3: When multiple interpretations of a command are detected, the optimal solution is selected based on the node associations in the semantic graph; D4: Feedback of the resolution results to the voice interaction interface via a confidence score.

[0079] In this embodiment, it should be noted that after the multimodal data fusion module 401 integrates voice, text and environmental information, the dialogue state tracking module 402 establishes a temporal context, the user portrait modeling module 403 injects personalized preferences, and finally the dynamic semantic graph of the context association analysis module 404 is used for disambiguation.

[0080] In addition, it should be noted that the entities in D2 include devices, actions, and parameters. The specific formula for calculating the preference weight of a device is: Where N i is the number of times the device is used; t is the time since the last use, and α=0.6, β=0.4, and λ=0.1 are determined through experiments. Construction of dynamic semantic graph: ① Extract entities such as devices, actions, and parameters from the instruction text through the named entity recognition model, such as identifying the device entity "living room light" and the action entity "turn on" from "turn on the living room light"; ② Use dependency syntax analysis to determine the grammatical relationship between entities, and generate directed edges such as "action-device" and "parameter-value", such as "turn on-living room light" to represent the association between action and device; ③ Match the current entity with the cached historical entity, such as through the device name. If the match is successful, a cross-round relationship is established, such as "current action-historical action", otherwise an independent node is added; ④ Use the attribute graph model to store entities and relationships, add nodes / edges after each round of dialogue and update the timestamp, and generate entity embedding vectors through the graph neural network for subsequent semantic ambiguity resolution. The correlation calculation of the semantic graph is as follows: Score(e i ,e j )=w1·TF-IDF(e i ,e j )+w2·KG_Sim(e i ,e j ), where e i ,e j is the entity node; w1=0.7, w2=0.3 are the experimental optimization weights.

[0081] In the present invention, the control execution unit 500 includes a protocol conversion module 501, a device driver module 502, a state feedback module 503, and an error handling and retry module 504, wherein: the protocol conversion module 501 is used to convert the instructions parsed by the AI semantic analysis unit 400 into a control protocol supported by the target device, such as Bluetooth HID and infrared coding, to achieve adaptive conversion of multiple protocols; the device driver module 502 is used to generate a corresponding drive signal according to the converted control protocol to trigger the target device to perform corresponding operations; the state feedback module 503 is used to receive the state information of the target device and feed it back to the system so that the user can understand the execution status of the device; the error handling and retry module 504 is used to detect and handle possible errors during the control execution process, such as communication failures and abnormal device responses, and perform retry operations.

[0082] In this embodiment, it should be noted that the protocol conversion module 501 dynamically adapts the semantic instructions to the target device protocol, and generates a precise control signal through the device driver module 502 to trigger execution. At the same time, the status feedback module 503 monitors the device status in real time to form a closed-loop feedback, and the error handling and retry module 504 ensures operational reliability.

[0083] Throughout this specification, references to terms such as "one embodiment," "example," or "specific example" indicate that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of these terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0084] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent Bluetooth voice remote control system based on AI semantic analysis, characterized in that: The system comprises a voice collection unit (100), a Bluetooth communication unit (200), a voice recognition unit (300), an AI semantic analysis unit (400), and a control execution unit (500), wherein: The voice collection unit (100) is used to collect the user's voice instructions and perform pre-processing; The Bluetooth communication unit (200) is used to establish a low-power Bluetooth connection with a target device, supporting two-way data transmission; The speech recognition unit (300) is used to convert speech instructions into text through automatic speech recognition technology, supports multi-language and dialect adaptation, and improves the recognition accuracy and real-time performance of accents and professional terms through localized model training and edge computing optimization; The AI semantic analysis unit (400) is based on intelligent semantic analysis of multimodal fusion, and realizes intention understanding, ambiguity resolution and personalized interaction through dialogue state tracking, user portrait modeling and context association analysis; The control execution unit (500) is used to convert the parsed instruction into a control protocol supported by the target device, support multi-protocol adaptive conversion of Bluetooth HID and infrared coding, and trigger execution and status feedback.

2. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 1, characterized in that: The voice collection unit (100) comprises a microphone array module (101), a pre-processing module (102), a voice activity detection module (103), and a low-power wake-up module (104), wherein: the microphone array module (101) is used to collect voice signals emitted by a user from the surrounding environment, and can capture sounds from multiple directions and angles; The pre-processing module (102) is responsible for performing noise reduction, echo cancellation and voice enhancement processing on the captured voice signal; The voice activity detection (103) is used to identify valid voice segments and filter out silence and background noise; The low-power waking module (104) is used to wake up the device through keyword detection, thereby reducing standby power consumption.

3. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 1, characterized in that: The Bluetooth communication unit (200) comprises a Bluetooth chip module (201), a radio frequency front-end module (202), a Bluetooth protocol stack module (203), and a data cache and processing module (204), wherein: The Bluetooth chip module (201) is responsible for implementing the underlying functions of the Bluetooth protocol, including signal modulation and demodulation, and frequency switching; The radio frequency front-end module (202) is used to perform power amplification and filtering processing on the Bluetooth signal, thereby enhancing the signal's transmission capability and anti-interference capability; The Bluetooth protocol stack module (203) is used to implement various protocols and specifications of Bluetooth communication, such as pairing, connection management, and data transmission, and provide a standardized interface for upper layer applications; The data buffering and processing module (204) is used to temporarily store data to be sent and received, and perform necessary processing.

4. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 1, characterized in that: The speech recognition unit (300) includes a feature extraction module (301), a localization model library module (302), an edge computing module (303), and a multi-language adaptation module (304), wherein: The feature extraction module (301) is used to extract representative feature parameters, such as Mel-frequency cepstral coefficients, from the preprocessed speech signal; The localized model library module (302) is used to store speech recognition models trained for different regions, dialects and professional fields, and select an appropriate model for recognition according to the user's usage scenario; The edge computing module (303) is used to perform some computing tasks on the local device, reduce data transmission delay, improve the real-time performance of recognition, and protect user data privacy; The multi-language adaptation module (304) is used to support the recognition of multiple languages and dialects, and to achieve accurate recognition in different language environments by switching between language models and vocabularies.

5. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 4, characterized in that: The localized model library module (302) is used to store speech recognition models trained for different regions, dialects and professional fields, and selects a suitable model for recognition according to the user's usage scenario. The specific operations are as follows: A1: Automatically trigger model switching based on user location, connection device type, or time characteristics; A2: Load pre-trained dialect / domain models, such as Cantonese and medical terminology models, from local encrypted storage. Compress the model size to ≤50MB to adapt to edge devices. A3: When it is detected that the user has corrected the recognition result three times in a row, the output weights of the general model and the domain-specific model are automatically mixed; A4: Use digital signatures to verify the integrity and origin of model files and prevent unauthorized model loading.

6. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 5, characterized in that: The edge computing module (303) is used to perform some computing tasks on the local device, reduce data transmission delay, improve the real-time performance of recognition, and protect user data privacy. The specific operations are: B1: The speech recognition process is divided into acoustic feature extraction performed locally and semantic understanding processed in the cloud, where acoustic feature extraction is completed on the device side; B2: Remove biometric information from voice features, such as voiceprint features, before uploading feature data, and retain content-related parameters; B3: Adjust the distribution ratio of local and cloud computing tasks in real time based on network latency and computing load; B4: Stores the acoustic model and language model of high-frequency words, and quickly retrieves them through Bloom filters, making the local recognition response time less than 200ms.

7. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 1, characterized in that: The AI semantic analysis unit (400) includes a multimodal data fusion module (401), a dialogue state tracking module (402), a user portrait modeling module (403), and a context association analysis module (404), wherein: The multimodal data fusion module (401) is used to fuse data of multiple modalities such as voice, text, and image, and comprehensively analyze the user's intention and context information; The dialogue state tracking module (402) is used to record and track the historical information and current state of the dialogue, providing contextual support for accurately understanding the user's intention; The user portrait modeling module (403) is used to construct a user portrait based on the user's historical behavior, preferences and usage habits, and realize personalized semantic analysis and interaction; The context association analysis module (404) is used to analyze the association relationship between the current conversation and the context, eliminate semantic ambiguity, and accurately understand the user's true intention.

8. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 7, characterized in that: The user portrait modeling module (403) is used to build a user portrait based on the user's historical behavior, preferences and usage habits, and realize personalized semantic analysis and interaction. The specific operations are as follows: C1: Records the user's historical command sequence, device operation frequency, and time period distribution; C2: Dynamically calculate the user's preference weight for the device based on the number of command executions and the most recent usage time; C3: Adopts a federated learning framework to aggregate local encrypted behavior data and update user profiles weekly. C4: When there is ambiguity in user instructions, the device or parameter with the highest weight in the profile is prioritized for execution.

9. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 8, characterized in that: The context association analysis module (404) is used to analyze the association between the current conversation and the context, eliminate semantic ambiguity, and accurately understand the user's true intention. The specific operations are as follows: D1: stores the command texts of the last three rounds of conversations and the environmental parameters of the device status in real time; D2: Build a dynamic semantic graph based on the current instruction and entities in the cache context; D3: When multiple interpretations of an instruction are detected, the optimal solution is selected based on the node association of the semantic graph; D4: Feedback the resolution results to the voice interaction interface through confidence scoring.

10. The intelligent Bluetooth voice remote control system based on AI semantic analysis according to claim 1, characterized in that: The control execution unit (500) includes a protocol conversion module (501), a device driver module (502), a state feedback module (503), and an error handling and retry module (504), wherein: The protocol conversion module (501) is used to convert the instructions parsed by the AI semantic analysis unit (400) into a control protocol supported by the target device, such as Bluetooth HID and infrared coding, to achieve adaptive conversion of multiple protocols; The device driver module (502) is used to generate a corresponding driving signal according to the converted control protocol, and trigger the target device to perform a corresponding operation; The state feedback module (503) is used to receive the state information of the target device and feed it back to the system so that the user can understand the execution status of the device; The error handling and retry module (504) is used to detect and handle possible errors during the control execution process, such as communication failures and abnormal device responses, and perform retry operations.

Images