Voice interaction recognition analysis method and system based on Bluetooth headset

Abstract

The invention relates to the field of voice recognition and analysis, in particular to a voice interaction recognition and analysis method and system based on a Bluetooth headset. The method comprises the following steps: identifying environment audio information of a current wearer based on a Bluetooth earphone; carrying out noise space distribution modeling on the environment audio information of the current wearer, and constructing an environment noise real-time distribution field; carrying out noise distribution change trend analysis on the environment noise real-time distribution field, and carrying out dynamic audio receiving adaptation calculation to obtain an optimal audio adaptation receiving parameter; carrying out real-time interactive audio perception on the wearer according to the optimal audio adaptive receiving parameters, carrying out sound quality detail enhancement processing, and constructing a sound quality detail enhanced audio signal; and performing deep audio semantic analysis and instant semantic ambiguity analysis on the sound quality detail enhanced audio signal, and marking an audio segment of semantic ambiguity. According to the invention, accurate and efficient voice interaction recognition analysis is realized.

Description

Technical Field

[0001] The present invention relates to the field of speech recognition analysis, and in particular to a speech interaction recognition analysis method and system based on a Bluetooth headset. Background Art

[0002] With the rapid development of smart devices and people's increasing demand for convenient interactive experience, voice interaction technology has become one of the core functions of modern electronic products. As a convenient mobile audio device, Bluetooth headsets have been widely used in various scenarios such as smartphones, tablets, smart homes and car systems, and have become a common tool for users to perform voice interaction. Through Bluetooth headsets, users can realize multiple functions such as voice command input, music playback control, call functions, and smart assistant operation, which greatly facilitates daily life and work.

[0003] However, during long-term use of Bluetooth headsets, the recognition accuracy of voice interaction may be affected by multiple factors such as environmental noise, the wearer's language style, and the quality of device hardware. Especially in complex environments, such as noisy streets, in cars, or during exercise, traditional voice recognition methods often have difficulty accurately distinguishing user commands, resulting in misrecognition or no response. In addition, due to the size limitations and hardware performance constraints of Bluetooth headsets, the audio quality of their voice input is often lower than that of professional microphone devices, further affecting the performance of the voice recognition system.

[0004] Traditional voice interaction recognition methods mostly rely on the collaboration between devices and the cloud, and have limited adaptability in different environments. This method usually cannot automatically adjust the recognition strategy according to environmental changes in real time, and cannot perform flexible adaptive processing when noise occurs or the wearer's emotions change, resulting in a poor user experience. Therefore, the existing voice interaction recognition technology urgently needs a more intelligent, accurate and adaptable solution to cope with increasingly complex usage scenarios and changing voice input characteristics. Summary of the invention

[0005] In order to solve the above technical problems, the present invention proposes a voice interaction recognition and analysis method and system based on Bluetooth headset to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides a method for voice interaction recognition and analysis based on a Bluetooth headset, comprising the following steps: Step S1: identifying the current wearer's ambient audio information based on the Bluetooth headset; performing noise spatial distribution modeling on the current wearer's ambient audio information to construct a real-time distribution field of ambient noise; Step S2: Analyze the noise distribution change trend of the real-time distribution field of the ambient noise, and perform dynamic audio reception adaptation calculation to obtain the optimal audio adaptation reception parameters; Step S3: performing real-time interactive audio perception of the wearer according to the optimal audio adaptation receiving parameters, and performing sound quality detail enhancement processing to construct a sound quality detail enhanced audio signal; Step S4: performing deep audio semantic analysis and real-time semantic ambiguity analysis on the audio signal with enhanced sound quality details, and marking audio segments with semantic ambiguity; Step S5: Obtain the wearer's historical audio log, perform conventional interaction logic evolution and personalized interaction habit learning, and build a deep interaction portrait of the wearer; Step S6: Perform precise adaptive semantic correction on the semantically ambiguous audio segments based on the wearer's deep interaction portrait, and then perform audio semantic recognition and reconstruction to obtain the speech interaction recognition result.

[0007] The present invention uses the built-in sensor and algorithm of the Bluetooth headset to obtain the ambient audio data around the wearer in real time and accurately model the noise distribution of the current environment. This process can help the system quickly adapt to the wearer's environment, identify and classify noise (such as wind, traffic, human voice, etc.), so that the headset can be flexibly adjusted in various noise environments. Through noise spatial distribution modeling, different noise sources can be monitored and their locations and intensities can be identified. Especially in complex or dynamically changing environments, this modeling method can continuously track the changes in noise sources and effectively enhance subsequent audio reception and processing capabilities. By analyzing the changing trend of the noise distribution field, changes in environmental noise can be predicted and adapted. This is especially important for wearers who are constantly moving or in an environment where different noise sources are intertwined. For example, in a quiet environment, the headset can select a high-sensitivity mode, and in a noisy environment, it will automatically switch to a noise cancellation mode. Through dynamic audio reception adaptation calculation, the system can provide optimal audio reception parameters for each environment and each wearer, ensuring the clarity and accuracy of voice signals in different environments. This adaptive optimization improves the wearer's voice interaction experience and reduces environmental interference. After adjusting the receiving parameters through optimal audio adaptation, the headset can capture the wearer's voice signal more accurately. Further sound quality detail enhancement processing can enhance the details in the voice signal, reduce audio distortion or fuzziness, and improve the accuracy of voice recognition, especially in complex voice environments. Sound quality detail enhancement not only improves the clarity of the audio, but also preserves the emotional changes and intonation in the voice, which is very beneficial for improving natural language understanding and emotional communication in voice interaction. For example, the system can better understand the wearer's tone and emotions, so as to respond more humanely. Through deep semantic analysis of audio signals, keywords, contexts and semantic levels in the voice can be mined, not just staying at the surface word level of recognition. This can better handle complex language structures and sentences and improve the accuracy of recognition results. Voice ambiguity is a common problem in complex voice commands or conversations. Through real-time semantic ambiguity analysis of audio signals, ambiguous paragraphs can be dynamically marked and identified, semantic deviations can be corrected in time, and incorrect recognition caused by ambiguity can be avoided, ensuring the intelligence and accurate response of the system. By analyzing the wearer's historical audio logs and interaction habits, the system can gradually understand the wearer's preferences and language habits. Based on this, personalized interaction logic evolution can be performed, and the response of the voice recognition system can be continuously optimized to make it more in line with the wearer's interaction method. By continuously accumulating the wearer's historical interaction data, the system will gradually form a deep interaction profile of the wearer. This adaptive learning mechanism allows the system to become smarter over time, able to predict and identify the wearer's future behavior and needs, and enhance the accuracy and efficiency of voice interaction. Based on the wearer's deep interaction profile, the system can accurately correct the ambiguous parts of the speech semantics to avoid misunderstandings caused by ambiguous words or unclear context.This enables the system to interpret user intent more effectively and respond accurately even in complex language environments. After correcting semantic ambiguity, the system will perform semantic reconstruction based on the corrected speech to improve the quality of the overall voice interaction. This process can eliminate unnecessary interference and errors, thereby generating high-quality interactive recognition results, making voice interaction more natural and smooth, and improving the user experience.

[0008] In this specification, a voice interaction recognition and analysis system based on a Bluetooth headset is provided, which is used to execute the voice interaction recognition and analysis method based on a Bluetooth headset as described above, including: A noise distribution module is used to identify the current wearer's ambient audio information based on the Bluetooth headset; perform noise spatial distribution modeling on the current wearer's ambient audio information to construct a real-time distribution field of ambient noise; The audio adaptation module is used to analyze the noise distribution change trend of the real-time distribution field of environmental noise, and perform dynamic audio reception adaptation calculation to obtain the optimal audio adaptation reception parameters; The sound quality enhancement module is used to perform real-time interactive audio perception of the wearer according to the optimal audio adaptation reception parameters, and to perform sound quality detail enhancement processing to construct a sound quality detail enhanced audio signal; A semantic analysis module is used to perform deep audio semantic analysis and real-time semantic ambiguity analysis on audio signals with enhanced sound quality details, and mark audio segments with semantic ambiguity; The interactive portrait module is used to obtain the wearer's historical audio logs, perform conventional interactive logic evolution and personalized interactive habit learning, and build a deep interactive portrait of the wearer; The semantic correction module is used to perform accurate and adaptive semantic correction on semantically ambiguous audio segments based on the wearer's deep interaction portrait, and then perform audio semantic recognition and reconstruction to obtain voice interaction recognition results.

[0009] The present invention uses the built-in microphone and sensor of the Bluetooth headset, and the noise distribution module can perceive the environmental noise (such as traffic noise, wind noise, human voice, etc.) around the wearer in real time, and identify the source, intensity and direction of the noise. This allows the system to adapt to dynamically changing environments and handle different types of noise sources. Through noise spatial distribution modeling, the system can locate the distribution and intensity of noise and provide more accurate environmental audio understanding. This provides reliable basic data for subsequent noise suppression and audio reception optimization. The audio adaptation module monitors the noise change trend in real time, predicts the noise changes that the wearer may encounter, and makes timely adjustments. Whether the wearer is in a static environment or a dynamic environment, the module can dynamically adjust the receiving strategy according to the noise changes. The module can automatically optimize the receiving parameters of the Bluetooth headset according to the wearer's real-time environment and interaction needs (for example, the wearer is in a quiet environment or a busy street) to ensure the clarity and accuracy of the voice signal, thereby improving the wearer's voice interaction experience. Through detail enhancement processing, the sound quality enhancement module can improve the clarity and details of the voice signal, especially in a noisy environment, to ensure that the wearer's voice input is clearly captured and conveyed. The enhancement processing can also filter out unnecessary environmental noise and highlight the true sound quality of the voice. This module not only improves the sound quality, but also can intelligently identify the emotions and intonation of the wearer's voice, allowing the system to respond more accurately to the wearer's emotional changes. For example, when the wearer's tone is excited, the system can optimize the voice signal processing to make its emotional communication clearer. The semantic analysis module can deeply analyze the voice content in the audio signal by combining deep learning and natural language processing technology. It can understand the contextual relationships, keywords and grammatical structures in the language to improve the accuracy of speech recognition. This module can identify semantic ambiguity in audio signals, especially in the case of homophones or polysemous words. The system can mark and annotate the suspected ambiguous parts in time to provide a basis for subsequent correction and avoid misunderstandings or recognition errors. The interactive portrait module can understand the wearer's language preferences, idioms and interaction patterns by analyzing the wearer's historical interaction logs. This enables the system to automatically adapt to the wearer's interaction needs in subsequent interactions and improve the personalization of the voice recognition system. This module continuously learns and adjusts based on the wearer's behavioral habits and interaction patterns, so that the system can respond more accurately to different voice commands and improve the user experience. Combined with the wearer's deep interaction profile, this module can accurately identify and correct semantic ambiguity in audio signals, reducing the possibility of misunderstanding. Through adaptive semantic correction, the system can accurately understand the wearer's intentions in complex contexts and avoid misrecognition caused by voice ambiguity or noise. Through semantic correction, the speech recognition system can more accurately understand the wearer's actual needs, thereby generating speech interaction results that are more in line with the wearer's intentions.This process improves the fluency and accuracy of speech recognition, ensuring that users can have a more natural and efficient voice interaction experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A schematic diagram of the steps of a method for speech interaction recognition and analysis based on a Bluetooth headset of the present invention; Figure 2 Detailed implementation flow chart of step S1; Figure 3 Detailed implementation flow chart of step S2; Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0011] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0012] The present application example provides a method and system for voice interaction recognition and analysis based on Bluetooth headsets. The execution subject of the method and system for voice interaction recognition and analysis based on Bluetooth headsets includes but is not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of the present application, and the data processing platform includes but is not limited to: at least one of audio image management system, information management system, and cloud data management system.

[0013] See also Figures 1 to 4 The present invention provides a voice interaction recognition and analysis method based on a Bluetooth headset, and the voice interaction recognition and analysis method based on a Bluetooth headset comprises the following steps: Step S1: identifying the current wearer's ambient audio information based on the Bluetooth headset; performing noise spatial distribution modeling on the current wearer's ambient audio information to construct a real-time distribution field of ambient noise; Step S2: Analyze the noise distribution change trend of the real-time distribution field of the ambient noise, and perform dynamic audio reception adaptation calculation to obtain the optimal audio adaptation reception parameters; Step S3: performing real-time interactive audio perception of the wearer according to the optimal audio adaptation receiving parameters, and performing sound quality detail enhancement processing to construct a sound quality detail enhanced audio signal; Step S4: performing deep audio semantic analysis and real-time semantic ambiguity analysis on the audio signal with enhanced sound quality details, and marking audio segments with semantic ambiguity; Step S5: Obtain the wearer's historical audio log, perform conventional interaction logic evolution and personalized interaction habit learning, and build a deep interaction portrait of the wearer; Step S6: Perform precise adaptive semantic correction on the semantically ambiguous audio segments based on the wearer's deep interaction portrait, and then perform audio semantic recognition and reconstruction to obtain the speech interaction recognition result.

[0014] The present invention uses the built-in sensor and algorithm of the Bluetooth headset to obtain the ambient audio data around the wearer in real time and accurately model the noise distribution of the current environment. This process can help the system quickly adapt to the wearer's environment, identify and classify noise (such as wind, traffic, human voice, etc.), so that the headset can be flexibly adjusted in various noise environments. Through noise spatial distribution modeling, different noise sources can be monitored and their locations and intensities can be identified. Especially in complex or dynamically changing environments, this modeling method can continuously track the changes in noise sources and effectively enhance subsequent audio reception and processing capabilities. By analyzing the changing trend of the noise distribution field, changes in environmental noise can be predicted and adapted. This is especially important for wearers who are constantly moving or in an environment where different noise sources are intertwined. For example, in a quiet environment, the headset can select a high-sensitivity mode, and in a noisy environment, it will automatically switch to a noise cancellation mode. Through dynamic audio reception adaptation calculation, the system can provide optimal audio reception parameters for each environment and each wearer, ensuring the clarity and accuracy of voice signals in different environments. This adaptive optimization improves the wearer's voice interaction experience and reduces environmental interference. After adjusting the receiving parameters through optimal audio adaptation, the headset can capture the wearer's voice signal more accurately. Further sound quality detail enhancement processing can enhance the details in the voice signal, reduce audio distortion or fuzziness, and improve the accuracy of voice recognition, especially in complex voice environments. Sound quality detail enhancement not only improves the clarity of the audio, but also preserves the emotional changes and intonation in the voice, which is very beneficial for improving natural language understanding and emotional communication in voice interaction. For example, the system can better understand the wearer's tone and emotions, so as to respond more humanely. Through deep semantic analysis of audio signals, keywords, contexts and semantic levels in the voice can be mined, not just staying at the surface word level of recognition. This can better handle complex language structures and sentences and improve the accuracy of recognition results. Voice ambiguity is a common problem in complex voice commands or conversations. Through real-time semantic ambiguity analysis of audio signals, ambiguous paragraphs can be dynamically marked and identified, semantic deviations can be corrected in time, and incorrect recognition caused by ambiguity can be avoided, ensuring the intelligence and accurate response of the system. By analyzing the wearer's historical audio logs and interaction habits, the system can gradually understand the wearer's preferences and language habits. Based on this, personalized interaction logic evolution can be performed, and the response of the voice recognition system can be continuously optimized to make it more in line with the wearer's interaction method. By continuously accumulating the wearer's historical interaction data, the system will gradually form a deep interaction profile of the wearer. This adaptive learning mechanism allows the system to become smarter over time, able to predict and identify the wearer's future behavior and needs, and enhance the accuracy and efficiency of voice interaction. Based on the wearer's deep interaction profile, the system can accurately correct the ambiguous parts of the speech semantics to avoid misunderstandings caused by ambiguous words or unclear context.This enables the system to interpret user intent more effectively and respond accurately even in complex language environments. After correcting semantic ambiguity, the system will perform semantic reconstruction based on the corrected speech to improve the quality of the overall voice interaction. This process can eliminate unnecessary interference and errors, thereby generating high-quality interactive recognition results, making voice interaction more natural and smooth, and improving the user experience.

[0015] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of the steps of a voice interaction recognition and analysis method based on a Bluetooth headset of the present invention. In this example, the steps of the voice interaction recognition and analysis method based on a Bluetooth headset include: Step S1: identifying the current wearer's ambient audio information based on the Bluetooth headset; performing noise spatial distribution modeling on the current wearer's ambient audio information to construct a real-time distribution field of ambient noise; In this embodiment, the built-in microphone of the Bluetooth headset is used to collect ambient audio. Ensure that the microphone of the headset is reasonably set and has moderate sensitivity to capture the surrounding sounds. It is recommended to set the sampling rate to 44.1 kHz to obtain a clear audio signal, and select a bit depth of 16 bits to ensure the dynamic range of the audio. During the audio acquisition process, the timestamp and duration of the collected audio are recorded. This information is very important for subsequent data analysis and helps to identify the changing trend of the ambient audio. It is recommended to record the ambient audio for at least 5 minutes to ensure the representativeness and sufficiency of the data. After the audio acquisition is completed, the original audio signal is preprocessed. This process includes removing silent segments, applying a high-pass filter to remove low-frequency noise (such as wind), and normalizing the volume to ensure the uniformity of the signal. The audio signal can be analyzed using a short-time Fourier transform (STFT) to obtain its spectral information. At this stage, the parameters used in the processing process, such as the cutoff frequency of the filter and the normalized target volume, are recorded. These parameters will provide basic data for subsequent noise spatial distribution modeling. Audio feature extraction is performed to identify the noise components in the ambient audio. There are many feature extraction methods that can be used, such as Mel frequency cepstral coefficients (MFCC), zero crossing rate (ZCR), and short-time energy. These features can help analyze the frequency distribution and changes of audio signals. Record the calculation method and parameter settings of each feature, such as the dimension and window length of MFCC. This information will play a key role in modeling the spatial distribution of noise. Use the extracted noise features to model the spatial distribution of noise. Select an appropriate modeling method, such as Kriging interpolation or inverse distance weighted (IDW), to generate a spatial distribution map of ambient noise. By combining the spatial coordinates of the audio signal (such as the GPS coordinates of the wearer's location) with the noise features, a real-time distribution field of ambient noise can be constructed. During the modeling process, make sure to record the parameter settings of each step, such as the weight of the interpolation and the spatial resolution of the model. This information is essential for optimizing the model and subsequent analysis. Finally, generate a visualization chart of the real-time distribution field of ambient noise to facilitate intuitive understanding of the noise distribution. Heat maps or three-dimensional graphs can be used to show the noise intensity in different areas to help identify noise sources and hot spots. Verify the results and compare the modeling results with the actual collected audio data to ensure the accuracy and reliability of the model. Record all parameters and results during the verification process for subsequent adjustment and optimization.

[0016] Step S2: Analyze the noise distribution change trend of the real-time distribution field of the ambient noise, and perform dynamic audio reception adaptation calculation to obtain the optimal audio adaptation reception parameters; In this embodiment, multiple noise monitoring sensors are deployed in the target area. These sensors should have high sensitivity and accuracy and be able to capture real-time noise data in different frequency ranges (such as 20Hz to 20kHz). In addition, ensure that the sensors have wireless transmission capabilities so that data can be transmitted to the central processing system in real time. The collected data should include noise intensity (expressed in decibels (dB)), spectrum information, timestamps, and sensor locations. All data should be stored in a standardized format (such as CSV or database) for subsequent analysis and processing. Data preprocessing: Before analysis, the collected environmental noise data is preprocessed, including removing outliers, smoothing, and normalization. This process can improve the quality of the data and ensure the accuracy of the analysis results. For example, the noise intensity data is smoothed using the moving average method. Select a suitable trend analysis method, such as time series analysis or spatial interpolation technology. Common time series analysis methods include autoregressive moving average (ARIMA) model and seasonal decomposition to identify the changing trend of noise distribution. Using the selected analysis method, the processed noise data is analyzed to identify the changing trend of noise intensity in different time periods. Generate trend graphs with peaks and valleys to identify noise sources and possible influencing factors. Adaptation parameter setting: Set the parameters required for dynamic audio reception adaptation calculation according to the trend of noise distribution changes. These parameters include receiving sensitivity, expected signal-to-noise ratio (target SNR should be set to above 20dB), and processing delay (ideally less than 50 milliseconds). Select appropriate dynamic adaptation algorithms, such as adaptive filtering or model-based optimization algorithms, to perform real-time audio reception adaptation calculations. These algorithms can dynamically adjust reception parameters according to the real-time noise environment to ensure the clarity and intelligibility of audio signals. Combine trend analysis results with set adaptation parameters to calculate optimal audio reception parameters in real time. By monitoring real-time noise levels, dynamically adjust receiving sensitivity and frequency response to optimize audio reception. For example, during periods of high noise intensity, enhance the filtering function of signal processing to reduce the impact of background noise. Effect evaluation: After each time period, evaluate the effect of dynamic audio reception and record the adapted audio quality indicators, including audio clarity, distortion, and user feedback. A comprehensive evaluation can be performed using both subjective evaluations (such as user satisfaction surveys) and objective indicators (such as signal-to-noise ratio).

[0017] Step S3: performing real-time interactive audio perception of the wearer according to the optimal audio adaptation receiving parameters, and performing sound quality detail enhancement processing to construct a sound quality detail enhanced audio signal; In this embodiment, the wearer's voice and environmental sound are recorded in real time using a built-in microphone. Ensure that the microphone has moderate sensitivity to capture the desired audio signal while avoiding excessive background noise interference. Before the audio signal is transmitted to the processing unit, preliminary signal preprocessing is performed. This includes applying a high-pass filter to remove low-frequency noise (such as wind noise) and using automatic gain control (AGC) to adjust the amplitude of the audio signal to ensure that the signal is within an acceptable range. The audio signal is spectrally analyzed by fast Fourier transform (FFT) to identify the energy distribution of different frequency components. Focus on the mid- and high-frequency bands (such as 3 kHz to 12 kHz) because these frequency bands usually contain key information of speech. Apply a dynamic range compression algorithm to adjust the dynamic range of the audio signal to make smaller sounds clearer while preventing larger sounds from causing distortion. This process can set thresholds and ratios to ensure that the processed audio sounds more balanced. Use an equalizer to fine-tune the spectrum and enhance the loudness of specific frequency bands. For example, the frequencies from 3 kHz to 6 kHz can be boosted to enhance the clarity of speech, while the low-frequency bands (such as below 100 Hz) can be appropriately reduced to reduce the interference of bass. The processed signal should have higher voice clarity, and the wearer can more easily hear voice commands and ambient sounds. After the sound quality details are enhanced, the dynamic range of the audio should be kept within a reasonable range, so that the audio signal can maintain good audibility in different environments. The signal after compression and equalization processing should avoid distortion as much as possible to ensure the naturalness and realism of the sound quality.

[0018] Step S4: performing deep audio semantic analysis and real-time semantic ambiguity analysis on the audio signal with enhanced sound quality details, and marking audio segments with semantic ambiguity; In this embodiment, a suitable deep learning model is selected, such as BERT or GPT, which performs well in natural language processing and can understand the grammatical structure and contextual information of sentences. If a pre-trained model is used, the extracted text can be directly input into the model for analysis. The model will output the deep semantic features of each audio segment, including information such as the theme, intent, and emotion. For example, the model may identify the theme of a certain audio segment as "weather advisory" and give a positive sentiment score. The parsing results are stored in a database to ensure that the deep semantic features of each audio segment are complete and traceable. Record the theme, sentiment score, intent and other information of each audio segment. Set the criteria for ambiguity analysis, including word sense ambiguity and syntactic ambiguity. For example, the word "bank" may refer to "financial institution" or "river bank" in different contexts. Combine the context information for analysis and evaluate the multiple meanings of each keyword or phrase. Use a context window (such as 5 words before and after) to help determine its specific meaning. Analyze the semantic ambiguity in the audio text with the help of a context-based model (such as ELMo or Transformers). The model encodes the context of each word to help identify possible ambiguous parts. Mark the identified semantically ambiguous segments as "ambiguous" or "clear" and record their corresponding timestamps. This not only helps with subsequent analysis, but also provides feedback to the wearer to help them understand potential misunderstandings. Generate a report containing the results of the ambiguous analysis, detailing the content, sentiment, theme, and tagging results for each audio segment. This report will provide users with a clear interaction review to facilitate understanding of potential problems in voice interaction. Store the results of deep audio semantic parsing and instant semantic ambiguity analysis in a database to ensure data integrity and traceability. Record the analysis results, tags, and related features of each audio segment for subsequent use.

[0019] Step S5: Obtain the wearer's historical audio log, perform conventional interaction logic evolution and personalized interaction habit learning, and build a deep interaction portrait of the wearer; In this embodiment, a suitable time range (such as the past month or longer) is selected to collect the wearer's historical audio logs. These logs should include all interaction records, such as voice commands, responses, and environmental noise. Ensure that all data complies with the privacy policy and obtain the wearer's consent for use. Store the collected audio logs in a structured database, with fields including audio duration, timestamp, voice content, device status, etc. Each record should be easy to query for subsequent analysis. Before analysis, the historical audio logs are preprocessed, including denoising, format conversion (such as converting audio to text), and tagging. The goal of preprocessing is to ensure the quality and consistency of the data and ensure the accuracy of subsequent analysis. Use a state machine model or a graph theory model to construct the wearer's interaction logic diagram. Each state represents an interaction scenario, and the transition between states represents the change of interaction. Set the transition condition, such as "After the wearer says 'play music', the system transfers to the 'music playing' state." Perform frequency analysis on each interactive logic state and count the wearer's interactive behavior in different scenarios. For example, record the frequency of occurrence of the "play music" state and the frequency of transition to its subsequent states. The analysis results should be recorded in the database for subsequent visualization and in-depth analysis. Extract key features from the evolution of interaction logic, such as the average duration of interaction, the stability of state transitions, etc. These features will help identify the wearer's normalized interaction patterns and provide a basis for personalized learning. Select appropriate machine learning models, such as cluster analysis (K-means) or decision trees, to learn and classify the wearer's interaction behaviors. Set the input features for learning, including keywords, frequency, and duration of historical interactions. Use historical audio logs for model training to identify the wearer's personalized interaction habits. For example, the model may identify that the wearer is more likely to ask about the weather in the morning and more likely to use music playback instructions in the evening. Record the frequency of occurrence of each habit and the corresponding contextual information. Store the learned personalized interaction habit features in the database to ensure data integrity and traceability. Record the features, frequency, and associated information of each habit for subsequent verification and optimization. Integrate all data obtained from logic evolution and personalized learning to build a deep interaction profile of the wearer. This profile should include the wearer's basic interaction features, behavior patterns, preferences, and contextual information. Use data visualization tools to generate a visualization chart of the wearer's interaction profile to display its interaction features and habits. These charts should be able to clearly show the wearer's interaction tendencies and changing patterns, making them easy to understand and analyze. Feedback the constructed deep interaction portraits to the wearer to provide personalized suggestions and optimization solutions. For example, inform the wearer of their common voice commands and the time periods when they interact most frequently to help them use the device more effectively.

[0020] Step S6: Perform precise adaptive semantic correction on the semantically ambiguous audio segments based on the wearer's deep interaction portrait, and then perform audio semantic recognition and reconstruction to obtain the speech interaction recognition result.

[0021] In this embodiment, the information of all previously marked ambiguous audio segments is extracted, including timestamps, identification text, marked ambiguity types and contexts. This process helps to understand the importance of each audio segment in a specific interaction. The features of each ambiguous audio segment, such as speech rate, pitch and emotional color, are analyzed to identify key factors that affect semantic understanding. These features can help the subsequent correction process be more targeted. Based on the wearer's deep interaction portrait, its common sentences and preferences are analyzed to determine the possible correct semantics of the audio segment. For example, if the wearer often asks about the weather in the morning, the understanding of the word "weather" should give priority to "weather forecast" rather than other meanings. An adaptive algorithm (such as a context-based decision tree or deep learning model) is applied to semantically correct the marked ambiguous segments. The model will input the current audio segment and its context information and output the possible correct semantics. For example, if "bank" is recognized, the model needs to determine whether it should be "financial institution" or "river bank" in combination with the context. The correction results are verified to ensure that the output semantics conform to the wearer's interaction habits. The effectiveness of the correction can be judged by setting a similarity threshold (e.g., 70%) to ensure that the corrected result is consistent with the wearer's deep interaction profile. Select a suitable audio reconstruction model, such as a sequence-to-sequence model (Seq2Seq) based on deep learning, to convert the corrected semantic information into understandable voice commands. Pass the corrected semantic information and its contextual features as input to the reconstruction model. Ensure that the model can handle a variety of input features, such as contextual keywords, sentiment tags, and interaction history, to improve the accuracy of the reconstruction. Generate new audio recognition results through the reconstruction model. Real-time monitoring should be carried out during this process to ensure that the output voice commands meet expectations and can be effectively recognized in different scenarios. Set evaluation criteria such as accuracy, recall, and F1 score to comprehensively evaluate the recognition results. The goal is to ensure that the recognition accuracy reaches more than 90%. If conditions permit, provide feedback to the wearer on the recognition results and provide a confirmation mechanism. The user can confirm whether the system recognizes the content accurately through simple voice feedback, such as "yes" or "no". This real-time feedback helps to further optimize the model. The final voice interaction recognition results and their evaluation data are stored in a database for subsequent analysis and model optimization. Record the parameter settings and performance during each recognition process to help optimize future interactive experiences.

[0022] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Step S11: Identifying the current wearer's ambient audio information based on the Bluetooth headset; Step S12: performing multi-time-frequency decomposition on the current wearer's ambient audio information to obtain multi-scale frequency domain data of the ambient audio; Step S13: identifying ambient audio noise points according to the ambient audio multi-scale frequency domain data, and marking a plurality of ambient noise points; Step S14: spatially locating multiple environmental noise points to obtain the position coordinates of each noise point; Step S15: Perform noise spatial distribution modeling according to the position coordinates of each noise point to construct a real-time distribution field of environmental noise.

[0023] In this embodiment, ensure that the Bluetooth headset has a high-quality microphone and audio processing function. Select a headset with noise reduction function to improve the accuracy of capturing ambient audio. Check whether the Bluetooth connection of the headset is stable to ensure that audio data can be effectively transmitted throughout the test process. Configure the audio acquisition parameters of the headset, including sampling rate and bit depth. Generally, a sampling rate of 44.1 kHz and a depth of 16 bits are selected to obtain better sound quality and details. In the environment where the wearer is normally active, start the audio acquisition function of the headset. Continuously collect ambient audio information, and it is recommended that the collection time be at least 5-10 minutes to ensure that sufficient audio data is captured to cover different ambient sounds. While recording audio data, collect the wearer's location information (such as GPS data) so that the audio information can be related to the specific environment during subsequent analysis. During the collection process, the ambient audio is preliminarily processed using the built-in signal processor of the headset to remove background noise and enhance useful audio signals. Apply an adaptive filtering algorithm to improve the clarity and audibility of the ambient audio. Transmit the processed audio data to a mobile phone or computing device for subsequent analysis to ensure that the data is complete and immediately available. Use wavelet transform or short-time Fourier transform (STFT) to perform multi-time-frequency decomposition on the captured ambient audio data. These methods can effectively extract the time-frequency features in the audio signal and are suitable for the analysis of non-stationary signals. Determine the window length and overlap rate of the wavelet basis function or Fourier transform. The window length is usually selected as 256ms and the overlap rate is 50% to obtain higher time resolution and frequency resolution.

[0024] Perform time-frequency decomposition on the audio data to extract the frequency components and amplitude information at each time point. Wavelet transform can be used to obtain the representation of the audio signal at different frequency scales, thereby obtaining multi-scale frequency domain data. Record the energy distribution of each time period and frequency range, generate a time-frequency diagram, and convert it into multi-scale frequency domain data. These data will be used for subsequent noise point identification. Store the multi-scale frequency domain data in a database to ensure that each time point, frequency and its corresponding energy value are recorded. You can use CSV file format or database system for storage. Generate a visual display of the time-frequency diagram to make it easy to understand and analyze. Displaying the energy distribution of different frequencies through a chart can help identify potential noise frequency ranges. The range is 200Hz to 2000Hz, and the energy threshold is a certain quantile (such as 90%) to facilitate the identification of significant noise points. Select an appropriate algorithm, such as a threshold-based detection method or a machine learning algorithm (such as a support vector machine) for noise point identification to ensure that environmental noise and background sound can be effectively distinguished. Analyze the multi-scale frequency domain data and mark the frequency components that are higher than the set threshold. By traversing the frequency domain data, prominent noise points are identified and their timestamps and frequency information are recorded. A list of noise points is generated, and the characteristics of each noise point are recorded, including the time of occurrence, frequency, and energy value. These data will be used for subsequent spatial positioning. The identified noise point data is stored in a database to ensure that the information of each noise point is complete and traceable. The recorded content includes timestamps, frequencies, energy values, and noise types (such as traffic noise, conversations, etc.). Determine the parameters required for spatial positioning of noise points, including the wearer's geographic location, audio signal propagation speed, and environmental conditions (such as temperature and humidity). These parameters affect the accuracy of sound propagation and the positioning accuracy of noise points.

[0025] Select an appropriate positioning method, such as a sound source localization algorithm (such as TDOA or AOA), so that the spatial position of the noise point can be determined based on the sound signal of the noise point. Use the recorded timestamp and frequency information of the noise point, combined with the wearer's location information, to calculate the spatial coordinates of each noise point. Use the sound source localization algorithm to analyze the arrival time difference or angle of the audio signal to deduce the location of the noise source. Record the coordinate information of each noise point, including the X, Y, and Z three-dimensional coordinates, to facilitate subsequent spatial distribution modeling. Store the located noise point coordinates in the database to ensure data integrity and traceability. Each record should contain the timestamp, frequency, energy value and spatial coordinates of the noise point. Determine the type of noise spatial distribution model, usually a grid-based model or a Kriging interpolation model, so as to construct a noise distribution field in three-dimensional space. Set the spatial resolution of the model, for example, choose the size of each grid to be 1 meter. Select appropriate model parameters, such as the noise attenuation coefficient, and consider the impact of environmental factors (such as obstacles, spatial structures) on sound wave propagation. Based on the stored noise point coordinate data, use modeling tools (such as MATLAB, Python, etc.) to build a noise spatial distribution model. Combine the intensity information of each noise point with its spatial coordinates to generate a preliminary model of the noise distribution field. Perform model optimization, consider the propagation characteristics of sound waves in space, and adjust the intensity and range of noise distribution. For example, weighted average or interpolation methods can be used to make the noise distribution in the model smoother and more realistic. Store the constructed noise spatial distribution model in a database to ensure the integrity and traceability of the model. Record the parameter settings and calculation results of the model for subsequent analysis and optimization. Generate a visualization of the noise spatial distribution, including a heat map or a three-dimensional graph, to show the noise intensity distribution in different areas. Through visualization, it is easy to quickly identify the hot spots of noise sources and provide data support for subsequent noise control measures.

[0026] In this embodiment, the specific steps of step S15 are: Perform environmental noise distribution analysis based on the position coordinates of each noise point to generate environmental noise distribution characteristics; Perform noise spectrum component analysis on multiple environmental noise points to obtain spectrum component information of each noise point; Spectrum envelope feature recognition is performed based on the spectrum component information of each noise point to generate spectrum envelope features of each noise point; Perform noise frequency time series change analysis on the spectrum component information of each noise point to generate noise frequency time series change data; Perform environmental noise point type analysis based on noise frequency time series change data and a noise point spectrum envelope feature to obtain the type of each noise point; Calculate the frequency intensity of multiple environmental noise points one by one to generate the frequency intensity of each noise point; Based on the frequency intensity of each noise point and the type of each noise point, the noise spatial distribution characteristics of the environmental noise are modeled to construct the real-time distribution field of environmental noise.

[0027] In this embodiment, the location information of the noise points is collected, including the three-dimensional coordinates (X, Y, Z) of each noise point. Ensure that these data are combined with the noise intensity and spectral component information for comprehensive analysis. Import the data into spatial analysis software (such as ArcGIS or QGIS) to prepare for subsequent distribution analysis. Use spatial interpolation techniques (such as Kriging interpolation or inverse distance weighted method) to analyze the noise points and generate a spatial distribution model of environmental noise. These methods can infer the noise distribution at unknown locations based on the intensity and location of known noise points. Generate a heat map of noise distribution to show the noise intensity in different areas and help identify noise hotspots and potential sources of interference. Visualize the generated noise distribution heat map, mark areas with high noise intensity, and analyze them in detail. Identify which areas of noise may affect the living or working environment of residents, and make corresponding management suggestions. Record the noise distribution characteristics of each area, including noise intensity, range and potential impact, for subsequent monitoring and control. Collect audio data from each noise point and ensure that the collection quality of these data meets the requirements, such as the sampling rate should be 44.1 kHz or higher to ensure the accuracy of spectral analysis. Import the audio data into a spectrum analysis tool (such as MATLAB, Audacity, or Python's Librosa library) for processing. Use Fast Fourier Transform (FFT) on the audio signal of each noise point to convert the time domain signal into a frequency domain signal. Record the spectral components of each noise point, including the energy distribution in different frequency ranges. Generate a spectrogram to display the frequency components and their intensity of each noise point, and analyze the frequency characteristics of different noise sources to help identify the type and source of noise. Store the spectral component information of each noise point in a database, including frequency, intensity, and its corresponding noise point information, to ensure data integrity and traceability. Perform a comparative analysis of the spectral components to identify the differences between different noise sources and record possible sources (such as traffic, industry, etc.). Determine the method for extracting the envelope features of the spectrum. Commonly used techniques include Hilbert transform or peak detection algorithm to extract the envelope features of the spectrum. Set the parameters for feature extraction, including the window length and overlap rate of envelope calculation, to ensure the accuracy of the features. Apply Hilbert transform to the spectral data of each noise point to extract the envelope features of its spectrum. Record the envelope features of each noise point, including the highest point, the lowest point, and its changing trend. Generate an envelope diagram to display the spectral envelope characteristics of each noise point for subsequent analysis and comparison. Store the extracted spectral envelope characteristics in a database to ensure that the characteristic information of each noise point is complete and traceable. Record the feature extraction process and parameter settings for subsequent verification and analysis. Perform comparative analysis of envelope characteristics to identify the characteristic differences of different noise sources and explore their relationship with noise type and spectral components. Collect time-series audio data for each noise point to ensure that the data timestamp is accurate and that each data point can reflect real-time noise changes.Import the timing data into the analysis tool to prepare for frequency timing change analysis.

[0028] Perform short-time Fourier transform (STFT) on the audio signal of each noise point to obtain spectrum information at different time points and record the frequency changes over time. Generate a frequency time series diagram to show the frequency changes of each noise point in different time periods and identify the frequency fluctuation pattern and trend. Store the frequency time series change data in the database to ensure that the time series information of each noise point is complete and traceable. Record the parameters and methods used in the analysis process for subsequent review and verification. Perform a comparative analysis of time series changes to identify the noise change trend in different time periods and explore its relationship with environmental factors (such as weather, traffic flow, etc.). Determine the classification criteria for noise points and conduct multi-dimensional analysis based on spectrum characteristics, time series changes and environmental factors. Noise types can be classified into traffic noise, industrial noise, environmental noise, etc. Select a suitable classification algorithm, such as KNN (K nearest neighbor) or support vector machine (SVM), to facilitate the automatic classification of noise points. Use the set classification method to analyze the data of each noise point and determine the type of each noise point by combining the spectrum envelope characteristics and frequency time series change data. Record the classification results of each noise point, including the selection of classification algorithms, parameter settings and output results, to ensure the transparency of the classification process. Store the noise type analysis results in the database to ensure that the type information of each noise point is complete and traceable. Record the parameter settings and results during the classification process for subsequent analysis and verification. Analyze the classification results to identify the characteristics and distribution of different types of noise, and explore their impact on the environment and potential management countermeasures. Determine the calculation criteria for frequency intensity, including the frequency range and weight of the calculation. The frequency range usually selected is 20Hz to 20kHz to cover all frequencies audible to the human ear. Set the calculation method, such as using root mean square (RMS) calculation or energy calculation, to accurately evaluate the frequency intensity of each noise point. Perform intensity calculation on the spectrum data of each noise point and output the intensity information of each frequency point. Record the results of the intensity calculation, including the calculation method and parameter settings. Generate a frequency intensity graph to show the frequency intensity distribution of each noise point to help identify the main components of the noise. Store the frequency intensity data of each noise point in the database to ensure the integrity and traceability of the data. Record the parameter settings and results during the calculation process for subsequent review and analysis. Conduct comparative analysis of frequency intensity, identify intensity differences among different noise sources, and explore their impact on the environment and management strategies.

[0029] Determine the type of noise spatial distribution model. Commonly used models include grid-based models or Kriging interpolation models. Select an appropriate spatial resolution, such as setting the size of each grid to 1 meter. Set the parameters of the model, including the noise attenuation coefficient, and consider the impact of environmental factors (such as terrain and buildings) on sound wave propagation. Use modeling tools (such as MATLAB, Python, etc.) to build a noise spatial distribution model based on the frequency intensity and type information of each noise point. Combine the intensity information of each noise point with its spatial coordinates to generate a preliminary model of the noise distribution field. Perform model optimization, consider the propagation characteristics of sound waves in space, and adjust the intensity and range of noise distribution to ensure the accuracy of the model. Store the constructed noise spatial distribution model in the database to ensure the integrity and traceability of the model. Record the parameter settings and calculation results of the model for subsequent analysis and optimization. Generate a visualization of the noise spatial distribution, including a heat map or a three-dimensional graph, to show the noise intensity distribution in different areas. Through visualization, it is easy to quickly identify the hot spots of noise sources and provide data support for subsequent noise management.

[0030] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Step S21: Analyze the noise distribution change trend of the real-time distribution field of environmental noise to generate noise distribution change trend characteristics; Step S22: performing self-learning of the noise distribution change trend characteristics to generate a dynamic change law of the distribution field; Step S23: Performing real-time scene analysis on the wearer based on the dynamic change rule of the distribution field, thereby generating a real-time audio scene; Step S24: performing scene audio trend prediction on the real-time audio scene to generate scene audio trend prediction data; Step S25: Perform dynamic audio reception adaptation calculation based on the scene audio trend prediction data to obtain optimal audio adaptation reception parameters.

[0031] In this embodiment, real-time environmental noise distribution data is collected, including noise intensity, frequency components and their spatial locations. Ensure that the timestamp of the data is accurate so that the noise changes in different time periods can be analyzed. Store the data in a database for subsequent query and analysis. Each record should include information such as time, noise intensity, frequency and spatial coordinates. Select a suitable time series analysis method, such as moving average, exponential smoothing or ARIMA (autoregressive integrated moving average) model, to identify the changing trend of noise distribution. Determine the time window for analysis, such as in hours or days, to facilitate capturing short-term and long-term changing trends. Perform trend analysis on the collected noise data and calculate the changes in noise intensity in different time periods. Record the mean, variance and standard deviation of each time period to assess the volatility of the noise level. Generate a noise distribution change trend report, showing a trend graph of noise intensity over time, so as to identify peak and trough periods. Store the results of the change trend analysis in the database to ensure the integrity and traceability of the data. Include the analysis results, noise level and its changing trend for each time period. Select a suitable machine learning model for self-learning analysis, such as recurrent neural network (RNN) or long short-term memory network (LSTM), which can effectively capture the changing trend of time series data. Determine the input features of the model, including noise intensity, frequency components and their timestamps, and set appropriate training parameters (such as learning rate, batch size, etc.). Organize the noise distribution change trend feature data into a format suitable for model input, such as using sliding window technology to split the time series data into training samples. Train the model, use historical noise distribution data for learning, and optimize model parameters to improve prediction accuracy. Cross-validation can be used to evaluate model performance. Use the trained model for real-time data analysis to identify the current noise distribution change trend and predict future changes. Record the output results of the model, including the predicted noise intensity and change trend. Generate a dynamic change law report to display the self-learning results of noise distribution to help understand the change law of noise in different scenarios. Collect real-time data of the wearer in a specific environment, including audio collection, location information and activity status. This data will be used to analyze the audio scene in which the wearer is located. Ensure the quality of audio data, use a microphone with a high sampling rate for recording, and record real-time changes in environmental audio. Select an appropriate audio scene recognition method, such as a feature extraction-based classification algorithm (such as a support vector machine or deep learning model) to classify and analyze real-time audio data. Determine the features for analysis, including spectral features, energy distribution, and timing changes, to identify the type of audio scene the wearer is in. Analyze the wearer's real-time audio data, extract audio features, and apply the selected classification model to identify the wearer's current audio scene type (such as quiet, noisy, traffic, etc.). Record the analysis results, including the type and features of the real-time audio scene and its corresponding timestamp.Select appropriate time series prediction models, such as ARIMA models, LSTM, or regression-based models, to predict the changing trends of real-time audio scenes. Determine the input features of the model, including historical audio scene data, noise intensity, and frequency components, to make effective trend predictions. Organize the collected real-time audio scene data into a format suitable for model input and perform data preprocessing, such as standardization and missing value processing. Train the prediction model, use historical audio scene data for learning, and optimize model parameters to improve the accuracy of predictions. Use cross-validation to evaluate model performance. Use the trained model to predict the trend of real-time audio scenes and record the audio changes in the future time period, including the expected changes in noise intensity and frequency. Generate an audio trend prediction report to show the changing trends of future audio scenes to facilitate corresponding preparation and management. Determine the parameters required for audio reception adaptation calculations, including audio quality, latency requirements, and reception sensitivity, to ensure the accuracy of adaptation calculations. Set adaptation strategies, such as adjusting reception parameters according to different audio scene types to optimize user experience. Perform dynamic audio reception adaptation calculations based on the predicted audio trend data and the set adaptation parameters. Record the results of the adaptation calculation, including the optimized reception parameter settings. Evaluate the impact of different reception parameters on audio quality to ensure the best audio reception effect in different scenarios. Store the results of dynamic audio reception adaptation calculation in the database to ensure data integrity and traceability, including the reception parameters and their performance in each scenario. Make subsequent optimization adjustments based on the reception effect and continuously improve the audio reception adaptation strategy. Generate a visual report to display the audio reception parameters and their effects in different scenarios to help users understand the changes in the adaptation strategy. Provide a user feedback mechanism to collect wearers' opinions on the audio reception effect in order to further optimize the adaptation calculation.

[0032] In this embodiment, refer to Figure 4 , is a flowchart of detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include: Step S31: performing real-time interactive audio perception of the wearer on the Bluetooth headset according to the optimal audio adaptation receiving parameters, and extracting the wearer's real-time input audio signal; Step S32: performing audio frequency offset recognition on the wearer's real-time input audio signal to identify the offset frequency segment; Step S33: performing offset compensation processing on the offset frequency segment to obtain a frequency offset optimized audio signal; Step S34: performing adaptive filtering and noise reduction on the frequency offset optimized audio signal to obtain a filtered optimized audio signal; Step S35: Perform sound quality detail enhancement processing on the filtered optimized audio signal to construct a sound quality detail enhanced audio signal.

[0033] In this embodiment, the audio input parameters of the Bluetooth headset are configured and set according to the optimal audio adaptation receiving parameters generated in the previous step. These parameters may include sampling rate (such as 44.1 kHz), bit depth (such as 16 bits) and number of channels (such as mono or stereo). The audio acquisition function of the headset is started to capture the audio signals around the wearer in real time. Ensure that the microphone sensitivity of the headset is moderate so that the ambient sound and the wearer's voice control input can be captured. The collected audio signal stream is transmitted to the processing unit (such as a mobile phone or a dedicated audio processing chip) in real time for subsequent analysis. Ensure that the data transmission delay is less than 50 milliseconds to ensure real-time performance. In the processing unit, the audio signal is pre-processed to remove obvious background noise and irrelevant audio signals to ensure the data quality of subsequent analysis. Implement a real-time monitoring mechanism to record the audio signal characteristics input by the wearer, including audio intensity, frequency range and duration, etc., for subsequent analysis. Provide users with a real-time feedback mechanism, such as through the prompt tone or vibration of the headset, so that the wearer knows whether his input audio is successfully captured to ensure the effectiveness of the interaction. Determine the detection standard for frequency offset and set an acceptable offset range (such as ±200Hz). This range should be set according to the frequency response characteristics of the headphones and the auditory characteristics of the wearer. Select appropriate frequency analysis tools, such as fast Fourier transform (FFT) or short-time Fourier transform (STFT), to perform frequency analysis on the input audio signal. Perform spectrum analysis on the real-time input audio signal to identify the offset frequency band. Record the intensity and phase information of each frequency point for subsequent compensation processing. Generate a frequency offset report, including the specific value of each offset frequency band and its corresponding energy intensity, to provide a basis for subsequent compensation processing. Store the identified offset frequency band information in the database to ensure data integrity and traceability. Record the parameter settings and results during the identification process for subsequent analysis and optimization. Analyze the offset frequency band and identify the possible causes of frequency offset, such as environmental noise or equipment failure, to provide information support for subsequent compensation and processing. Determine the appropriate frequency offset compensation algorithm. Common methods include phase compensation method and amplitude correction method. The selected algorithm should be able to effectively correct the identified offset frequency band. Set compensation parameters, including compensation amplitude and compensation frequency range, to ensure compensation accuracy. Perform compensation processing on the identified offset frequency bands, adjust the amplitude and phase information of the corresponding frequency bands, and ensure that the audio signal is restored to its original state. Record the compensation process, including the algorithms used, parameter settings, and their impact on the audio signal, for subsequent verification and review. Store the offset-compensated audio signal in a database to ensure data integrity and traceability. Record the processing results of each compensation frequency band for subsequent analysis. Verify the compensation effect and compare the audio signals before and after compensation to ensure the effectiveness and accuracy of the compensation.Select a suitable adaptive filtering algorithm (such as LMS algorithm, RLS algorithm) to perform noise reduction on the audio signal after frequency offset optimization. The selected algorithm should be able to dynamically adjust the filtering parameters to adapt to environmental changes. Determine the parameters of the filter, including the filter order and learning rate, to ensure the optimization of the filtering effect. Input the audio signal after frequency offset optimization into the adaptive filter, and process the audio signal in real time to remove background noise. Record the filtering effect of each time period, including indicators such as signal-to-noise ratio (SNR) and distortion rate. Generate a noise reduction processing result report to show the comparison of the audio signal before and after noise reduction to help identify the effect of noise reduction processing. Determine a suitable sound quality detail enhancement algorithm, commonly used methods include dynamic range compression, equalizer adjustment, and multi-band processing. The selected algorithm should be able to improve the detail performance of the audio signal. Set the enhancement parameters, including compression ratio, equalizer band gain, etc., to ensure the optimization of the enhancement effect. Apply the selected sound quality enhancement algorithm to the audio signal after filtering optimization to improve the detail performance and clarity of the audio signal. Record the enhancement process, including the setting of each parameter and its effect on the audio signal. Generate a sound quality enhancement processing result report to show the comparison of audio signals before and after processing, helping to identify the effect of sound quality improvement. Store the audio signal with enhanced sound quality details in the database to ensure data integrity and traceability. Record the sound quality enhancement process and parameter settings for subsequent verification and optimization. Generate a visual chart to show the audio waveform and spectrum before and after enhancement, helping to understand the specific effects and changes of sound quality improvement.

[0034] In this embodiment, step S4 includes the following steps: Step S41: Divide the audio signal with enhanced sound quality details into multiple time periods to generate multiple audio time segments; Step S42: performing keyword detection on multiple audio time segments one by one, and extracting keywords from each audio segment; Step S43: performing deep audio semantic analysis according to the keywords of each audio segment to obtain deep semantic features of each audio segment; Step S44: performing audio semantic logic recognition on the deep semantic features of each audio segment to generate a semantic logical structure of each audio segment; Step S45: Performing real-time semantic ambiguity analysis according to the semantic logical structure of each audio segment, and marking the semantically ambiguous audio segments.

[0035] In this embodiment, to ensure that the audio signal after the sound quality detail enhancement has good quality, the sampling rate should be maintained at 44.1 kHz or higher to ensure the clarity and details of the audio. The total duration of the audio signal is recorded for effective time division. Select an appropriate time segment length, and it is generally recommended that each segment length is 5 to 10 seconds for segment-by-segment analysis. Set the segment length according to the complexity of the audio content and the expected analysis depth. Use an audio processing tool (such as Python's Librosa library or MATLAB) to divide the enhanced audio signal. Divide the audio signal according to the set segment length to generate multiple audio time segments. During the segmentation process, ensure that each segment overlaps (such as 0.5 seconds) to retain the coherence and contextual information of the audio signal, which is convenient for subsequent keyword detection and semantic analysis. The generated audio time segments are stored in a database to ensure that the timestamp, duration and corresponding audio signal of each segment are complete and traceable. Record the feature information of each segment for subsequent analysis. Generate a segment segmentation report that describes the start and end time and content characteristics of each segment in detail to help identify the theme and importance of each segment. Model, so as to extract keywords from the audio signal. Determine the criteria for keyword extraction, including the keyword extraction threshold and frequency, to ensure that the extracted keywords are representative. Perform audio-to-text processing on each audio time segment, using speech recognition technology to convert the audio signal into text data. Ensure the use of a high-accuracy speech recognition model to reduce recognition errors. Apply the selected keyword detection algorithm to the text data to extract the keywords of each audio segment and record the frequency and position of each keyword. Store the extracted keywords and related information in the database to ensure that the source and characteristics of each keyword are complete and traceable. Record the keywords of each time segment and the corresponding audio segment information. Generate a keyword extraction report to display the keywords and their frequencies for each segment to help analyze the main themes and information of the audio content. Select a suitable deep semantic parsing model, such as a Transformer-based model or a pre-trained language model (such as BERT), to perform in-depth analysis of the extracted keywords. Determine the input features for parsing, including the keywords of each audio segment and their contextual information, to extract richer semantic information. Input the extracted keywords into the deep semantic parsing model to perform in-depth analysis of audio semantics to identify the theme, emotion, and potential meaning of each audio segment. Record the deep semantic features of each audio segment, including topic tags, sentiment scores (such as positive, negative, neutral), and related contextual information. Store the results of deep semantic analysis in a database to ensure that the semantic features of each audio segment are complete and traceable. Record the input and output of the model for subsequent verification and optimization. Generate a visual chart to display the semantic features of each audio segment to help quickly understand the audio content and its potential meaning.Determine the standards for audio semantic logic recognition, including the recognized logical relationships (such as causal relationships, contrast relationships, etc.) and logical structure types (such as declarative sentences, interrogative sentences, etc.). Select appropriate logic recognition algorithms, such as rule-based logical reasoning or machine learning models (such as decision trees, random forests), to perform semantic logical relationship analysis. Input the deep semantic features of each audio segment into the logic recognition model, analyze the logical relationships of the audio content, and identify its semantic logical structure. Record the semantic logic recognition results of each audio segment, including the recognized logical relationships, logical types, and their corresponding text information. Determine the standards for semantic ambiguity analysis, including ambiguity types (such as word sense ambiguity, syntactic ambiguity) and analysis methods (such as context analysis, contextual reasoning). Select appropriate analysis tools. Usually, NLP-based models (such as context-based language models) can be used for ambiguity recognition. Analyze the semantic logical structure of each audio segment and identify potential semantically ambiguous parts. Record each audio segment that may be ambiguous and its corresponding context information. Generate a semantic ambiguity report, which describes the ambiguous parts of each audio segment and their impact in detail to help with subsequent processing and optimization. The results of semantic ambiguity analysis are stored in a database to ensure that the ambiguous information of each audio segment is complete and traceable. The parameter settings and results during the analysis process are recorded for subsequent review. A feedback mechanism is provided to help users understand the ambiguous parts of the audio content and provide solutions or suggestions to enhance user understanding and experience.

[0036] In this embodiment, step S5 includes the following steps: Step S51: Obtain the wearer's historical audio log; Step S52: Perform historical interactive voice recognition on the wearer's historical audio log to extract historical interactive voice information; Step S53: performing conventional interaction logic evolution on the historical interaction voice information to obtain the wearer's normalized interaction logic law; Step S54: performing personalized interaction habit learning based on historical interaction voice information to obtain personalized interaction habit features; Step S55: Conduct deep interaction evolution on the wearer's normalized interaction logic rules and personalized interaction habit characteristics to construct a deep interaction portrait of the wearer.

[0037] In this embodiment, the scope and time period of data collection are determined, and a suitable historical audio log data source is selected. These data may include information such as the wearer's past audio interaction records, voice commands, and ambient audio. Ensure that the storage format of the historical audio log is available for subsequent analysis, usually in CSV or database format, with fields including audio duration, timestamp, voice content, etc. Collect the wearer's historical audio log through the recording function of Bluetooth headphones or mobile devices. Ensure that all audio materials comply with privacy policies and data protection regulations, and obtain the wearer's consent. Set the maximum sampling rate (such as 44.1 kHz) and bit depth (such as 16 bits) of the audio log to ensure audio quality and ensure the effectiveness of subsequent analysis. Store the acquired historical audio logs in a secure database to ensure data integrity and traceability. Record the source, time, and content characteristics of each log for subsequent analysis. Generate a summary report of the audio log, describing in detail the content and importance of each log, to help subsequent analysis identify the wearer's behavior pattern. Select a suitable speech recognition model, such as an automatic speech recognition (ASR) system based on deep learning, and ensure that it supports multiple accents and speech speeds to improve recognition accuracy. Determine the input parameters for recognition, including audio sampling rate, language model, and vocabulary, so as to effectively process historical audio logs. Input historical audio logs into the selected speech recognition model and convert them into text data in real time. Ensure that the recognition accuracy is monitored during the conversion process, and the target recognition accuracy should reach more than 90%. Record the recognition results of each audio segment, including the audio duration, recognized text, and its corresponding timestamp for subsequent analysis. Determine the analysis method for the evolution of conventional interaction logic. State machine models or graph theory models can be used to represent the relationship between different interaction states to model the logic evolution. Set the input parameters for analysis, including historical interaction voice information, recognized text, and its context information, so as to extract the interaction logic. Based on the recognized historical interaction voice information, build an interaction logic model to identify the wearer's interaction behavior and reaction patterns in different scenarios. Record the frequency and change trend of each interaction logic. Generate an interaction logic evolution report to show the evolution of conventional interaction logic and help identify the wearer's behavior patterns and habits. Select a suitable personalized learning algorithm, such as cluster analysis or machine learning model (such as random forest, K-means), to learn and analyze the wearer's interaction habits. Determine the input features of learning, including historical interactive voice information, recognition results and their contextual information, to extract personalized interaction habits. Analyze the historical interactive voice information using the selected learning algorithm to identify the wearer's personalized interaction habit features, such as common sentences, preferred requests and response patterns. Record the key features of the learning process, including the frequency of personalized habits, change trends and their corresponding voice commands.Select a suitable deep interaction portrait construction model. You can use a graph neural network (GNN) or a multi-layer perceptron (MLP) model to extract deep features from normalized interaction logic rules and personalized habit features. Determine the input features of the model, including normalized interaction logic and personalized habits, and set appropriate training parameters to improve model performance. Input the normalized interaction logic rules and personalized interaction habit features into the selected model, perform deep feature extraction and evolution analysis, and build a deep interaction portrait of the wearer. Record the construction process of the deep interaction portrait, including the training results of the model, key features and their impact on the wearer's behavior. Store the constructed deep interaction portrait in the database to ensure the integrity and traceability of the data. Record the parameter settings and results during the construction process to facilitate subsequent verification and optimization. Generate a visual chart to display the wearer's deep interaction portrait to help quickly understand the wearer's interaction characteristics and behavior patterns.

[0038] In this embodiment, step S6 includes the following steps: Step S61: inferring the wearer's interaction intention based on the wearer's deep interaction portrait and the keywords of each audio segment to generate the wearer's real-time interaction intention; Step S62: performing accurate adaptive semantic correction on the semantically ambiguous audio segment based on the wearer's real-time interaction intention to obtain a semantically corrected audio segment; Step S63: Perform audio semantic recognition and reconstruction on the semantically corrected audio segment and the deep semantic features of each audio segment, so as to obtain a speech interaction recognition result.

[0039] In this embodiment, a suitable interaction intention inference model is selected, such as a sequence-to-sequence model (Seq2Seq) or a Transformer model based on deep learning. These models can effectively use the wearer's deep interaction portrait and keyword information to infer intent. Determine the characteristics of the model input, including the wearer's deep interaction portrait, historical interaction keywords and their contextual information. Set appropriate training parameters to improve the accuracy of the model. Input the wearer's deep interaction portrait and the keywords of each audio segment into the selected intention inference model to infer the interaction intention. The model will analyze the relationship between the keywords and the interaction portrait to identify the wearer's intention in a specific context. Record each inference result, including the inferred intent category (such as query, command, feedback, etc.) and its confidence. Ensure the accuracy of each result for subsequent semantic correction. Store the inferred real-time interaction intention in the database to ensure the integrity and traceability of the data. Record the source of each intention inference (keywords and interaction portraits) for subsequent analysis. Generate an interaction intention inference report to display the inferred intent and its confidence for each audio segment to help understand the wearer's behavior patterns and needs. Determine the criteria for semantic correction, including identifying the types of semantic ambiguity (such as word sense ambiguity, syntactic ambiguity) and correction strategies (such as contextual replacement, synonym replacement, etc.). Select a suitable adaptive semantic correction algorithm, such as a context-based semantic understanding model or a rule-based correction method, to improve the accuracy of correction. According to the wearer's real-time interaction intention, analyze the identified semantically ambiguous audio segments and identify the parts that need to be corrected. Combined with the context information, apply the selected semantic correction algorithm for processing. Record each correction result, including the original audio segment, the corrected audio segment and its corresponding interaction intention, to ensure the traceability of the correction process. Select a suitable audio semantic recognition and reconstruction model, such as an audio classification model based on deep learning or a multimodal learning model, to perform semantic recognition on the corrected audio segment. Determine the input features of the model, including the semantically corrected audio segment and its deep semantic features, to ensure that the model can fully understand the audio content. Input the semantically corrected audio segment and its deep semantic features into the selected semantic recognition model to reconstruct and recognize the audio content. The model will analyze the relationship between the audio signal and the semantic features to generate the final speech interaction recognition result. Record each recognition result, including the recognized text, voice content and its corresponding confidence level, to facilitate subsequent analysis and optimization.

[0040] In this embodiment, a voice interaction recognition and analysis system based on a Bluetooth headset is provided, which is used to execute the voice interaction recognition and analysis method based on a Bluetooth headset as described above, including: A noise distribution module is used to identify the current wearer's ambient audio information based on the Bluetooth headset; perform noise spatial distribution modeling on the current wearer's ambient audio information to construct a real-time distribution field of ambient noise; The audio adaptation module is used to analyze the noise distribution change trend of the real-time distribution field of environmental noise, and perform dynamic audio reception adaptation calculation to obtain the optimal audio adaptation reception parameters; The sound quality enhancement module is used to perform real-time interactive audio perception of the wearer according to the optimal audio adaptation reception parameters, and to perform sound quality detail enhancement processing to construct a sound quality detail enhanced audio signal; A semantic analysis module is used to perform deep audio semantic analysis and real-time semantic ambiguity analysis on audio signals with enhanced sound quality details, and mark audio segments with semantic ambiguity; The interactive portrait module is used to obtain the wearer's historical audio logs, perform conventional interactive logic evolution and personalized interactive habit learning, and build a deep interactive portrait of the wearer; The semantic correction module is used to perform accurate and adaptive semantic correction on semantically ambiguous audio segments based on the wearer's deep interaction portrait, and then perform audio semantic recognition and reconstruction to obtain voice interaction recognition results.

[0041] The present invention uses the built-in microphone and sensor of the Bluetooth headset, and the noise distribution module can perceive the environmental noise (such as traffic noise, wind noise, human voice, etc.) around the wearer in real time, and identify the source, intensity and direction of the noise. This allows the system to adapt to dynamically changing environments and handle different types of noise sources. Through noise spatial distribution modeling, the system can locate the distribution and intensity of noise and provide more accurate environmental audio understanding. This provides reliable basic data for subsequent noise suppression and audio reception optimization. The audio adaptation module monitors the noise change trend in real time, predicts the noise changes that the wearer may encounter, and makes timely adjustments. Whether the wearer is in a static environment or a dynamic environment, the module can dynamically adjust the receiving strategy according to the noise changes. The module can automatically optimize the receiving parameters of the Bluetooth headset according to the wearer's real-time environment and interaction needs (for example, the wearer is in a quiet environment or a busy street) to ensure the clarity and accuracy of the voice signal, thereby improving the wearer's voice interaction experience. Through detail enhancement processing, the sound quality enhancement module can improve the clarity and details of the voice signal, especially in a noisy environment, to ensure that the wearer's voice input is clearly captured and conveyed. The enhancement processing can also filter out unnecessary environmental noise and highlight the true sound quality of the voice. This module not only improves the sound quality, but also can intelligently identify the emotions and intonation of the wearer's voice, allowing the system to respond more accurately to the wearer's emotional changes. For example, when the wearer's tone is excited, the system can optimize the voice signal processing to make its emotional communication clearer. The semantic analysis module can deeply analyze the voice content in the audio signal by combining deep learning and natural language processing technology. It can understand the contextual relationships, keywords and grammatical structures in the language to improve the accuracy of speech recognition. This module can identify semantic ambiguity in audio signals, especially in the case of homophones or polysemous words. The system can mark and annotate the suspected ambiguous parts in time to provide a basis for subsequent correction and avoid misunderstandings or recognition errors. The interactive portrait module can understand the wearer's language preferences, idioms and interaction patterns by analyzing the wearer's historical interaction logs. This enables the system to automatically adapt to the wearer's interaction needs in subsequent interactions and improve the personalization of the voice recognition system. This module continuously learns and adjusts based on the wearer's behavioral habits and interaction patterns, so that the system can respond more accurately to different voice commands and improve the user experience. Combined with the wearer's deep interaction profile, this module can accurately identify and correct semantic ambiguity in audio signals, reducing the possibility of misunderstanding. Through adaptive semantic correction, the system can accurately understand the wearer's intentions in complex contexts and avoid misrecognition caused by voice ambiguity or noise. Through semantic correction, the speech recognition system can more accurately understand the wearer's actual needs, thereby generating speech interaction results that are more in line with the wearer's intentions.This process improves the fluency and accuracy of speech recognition, ensuring that users can have a more natural and efficient voice interaction experience.

[0042] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0043] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A voice interaction recognition and analysis method based on Bluetooth headset, characterized in that: The following steps are involved: Step S1: identifying the current wearer's ambient audio information based on the Bluetooth headset; performing noise spatial distribution modeling on the current wearer's ambient audio information to construct a real-time distribution field of ambient noise; Step S2: Analyze the noise distribution change trend of the real-time distribution field of the ambient noise, and perform dynamic audio reception adaptation calculation to obtain the optimal audio adaptation reception parameters; Step S3: performing real-time interactive audio perception of the wearer according to the optimal audio adaptation receiving parameters, and performing sound quality detail enhancement processing to construct a sound quality detail enhanced audio signal; Step S4: performing deep audio semantic analysis and real-time semantic ambiguity analysis on the audio signal with enhanced sound quality details, and marking audio segments with semantic ambiguity; Step S5: Obtain the wearer's historical audio log, perform conventional interaction logic evolution and personalized interaction habit learning, and build a deep interaction portrait of the wearer; Step S6: Perform precise adaptive semantic correction on the semantically ambiguous audio segments based on the wearer's deep interaction portrait, and then perform audio semantic recognition and reconstruction to obtain the speech interaction recognition result.

2. The method for voice interaction recognition and analysis based on Bluetooth headset according to claim 1, characterized in that: The specific steps of step S1 are: Step S11: Identifying the current wearer's ambient audio information based on the Bluetooth headset; Step S12: performing multi-time-frequency decomposition on the current wearer's ambient audio information to obtain multi-scale frequency domain data of the ambient audio; Step S13: identifying ambient audio noise points according to the ambient audio multi-scale frequency domain data, and marking a plurality of ambient noise points; Step S14: spatially locating multiple environmental noise points to obtain the position coordinates of each noise point; Step S15: Perform noise spatial distribution modeling according to the position coordinates of each noise point to construct a real-time distribution field of environmental noise.

3. The method for voice interaction recognition and analysis based on Bluetooth headset according to claim 2, characterized in that: The specific steps of step S15 are: Perform environmental noise distribution analysis based on the position coordinates of each noise point to generate environmental noise distribution characteristics; Perform noise spectrum component analysis on multiple environmental noise points to obtain spectrum component information of each noise point; Spectrum envelope feature recognition is performed based on the spectrum component information of each noise point to generate spectrum envelope features of each noise point; Perform noise frequency time series change analysis on the spectrum component information of each noise point to generate noise frequency time series change data; Perform environmental noise point type analysis based on noise frequency time series change data and a noise point spectrum envelope feature to obtain the type of each noise point; Calculate the frequency intensity of multiple environmental noise points one by one to generate the frequency intensity of each noise point; Based on the frequency intensity of each noise point and the type of each noise point, the noise spatial distribution characteristics of the environmental noise are modeled to construct the real-time distribution field of environmental noise.

4. The method for voice interaction recognition and analysis based on Bluetooth headset according to claim 1, characterized in that: The specific steps of step S2 are: Step S21: Analyze the noise distribution change trend of the real-time distribution field of environmental noise to generate noise distribution change trend characteristics; Step S22: performing self-learning of the noise distribution change trend characteristics to generate a dynamic change law of the distribution field; Step S23: Performing real-time scene analysis on the wearer based on the dynamic change rule of the distribution field, thereby generating a real-time audio scene; Step S24: performing scene audio trend prediction on the real-time audio scene to generate scene audio trend prediction data; Step S25: Perform dynamic audio reception adaptation calculation based on the scene audio trend prediction data to obtain optimal audio adaptation reception parameters.

5. The method for voice interaction recognition and analysis based on Bluetooth headset according to claim 1, characterized in that: The specific steps of step S3 are: Step S31: performing real-time interactive audio perception of the wearer on the Bluetooth headset according to the optimal audio adaptation receiving parameters, and extracting the wearer's real-time input audio signal; Step S32: performing audio frequency offset recognition on the wearer's real-time input audio signal to identify the offset frequency segment; Step S33: performing offset compensation processing on the offset frequency segment to obtain a frequency offset optimized audio signal; Step S34: performing adaptive filtering and noise reduction on the frequency offset optimized audio signal to obtain a filtered optimized audio signal; Step S35: Perform sound quality detail enhancement processing on the filtered optimized audio signal to construct a sound quality detail enhanced audio signal.

6. The method for voice interaction recognition and analysis based on Bluetooth headset according to claim 1, characterized in that: The specific steps of step S4 are: Step S41: Divide the audio signal with enhanced sound quality details into multiple time periods to generate multiple audio time segments; Step S42: performing keyword detection on multiple audio time segments one by one, and extracting keywords from each audio segment; Step S43: performing deep audio semantic analysis according to the keywords of each audio segment to obtain deep semantic features of each audio segment; Step S44: performing audio semantic logic recognition on the deep semantic features of each audio segment to generate a semantic logical structure of each audio segment; Step S45: Performing real-time semantic ambiguity analysis according to the semantic logical structure of each audio segment, and marking the semantically ambiguous audio segments.

7. The method for voice interaction recognition and analysis based on Bluetooth headset according to claim 1, characterized in that: The specific steps of step S5 are: Step S51: Obtain the wearer's historical audio log; Step S52: Perform historical interactive voice recognition on the wearer's historical audio log to extract historical interactive voice information; Step S53: performing conventional interaction logic evolution on the historical interaction voice information to obtain the wearer's normalized interaction logic law; Step S54: performing personalized interaction habit learning based on historical interaction voice information to obtain personalized interaction habit features; Step S55: Conduct deep interaction evolution on the wearer's normalized interaction logic rules and personalized interaction habit characteristics to construct a deep interaction portrait of the wearer.

8. The method for voice interaction recognition and analysis based on Bluetooth headset according to claim 1, characterized in that: The specific steps of step S6 are: Step S61: inferring the wearer's interaction intention based on the wearer's deep interaction portrait and the keywords of each audio segment to generate the wearer's real-time interaction intention; Step S62: performing accurate adaptive semantic correction on the semantically ambiguous audio segment based on the wearer's real-time interaction intention to obtain a semantically corrected audio segment; Step S63: Perform audio semantic recognition and reconstruction on the semantically corrected audio segment and the deep semantic features of each audio segment, so as to obtain a speech interaction recognition result.

9. A voice interaction recognition and analysis system based on Bluetooth headset, characterized in that: The method for performing the voice interaction recognition and analysis method based on a Bluetooth headset as claimed in claim 1 comprises: A noise distribution module is used to identify the current wearer's ambient audio information based on the Bluetooth headset; perform noise spatial distribution modeling on the current wearer's ambient audio information to construct a real-time distribution field of ambient noise; The audio adaptation module is used to analyze the noise distribution change trend of the real-time distribution field of environmental noise, and perform dynamic audio reception adaptation calculation to obtain the optimal audio adaptation reception parameters; The sound quality enhancement module is used to perform real-time interactive audio perception of the wearer according to the optimal audio adaptation reception parameters, and to perform sound quality detail enhancement processing to construct a sound quality detail enhanced audio signal; A semantic analysis module is used to perform deep audio semantic analysis and real-time semantic ambiguity analysis on audio signals with enhanced sound quality details, and mark audio segments with semantic ambiguity; The interactive portrait module is used to obtain the wearer's historical audio logs, perform conventional interactive logic evolution and personalized interactive habit learning, and build a deep interactive portrait of the wearer; The semantic correction module is used to perform accurate and adaptive semantic correction on semantically ambiguous audio segments based on the wearer's deep interaction portrait, and then perform audio semantic recognition and reconstruction to obtain voice interaction recognition results.

Images