Intelligent communication system and equipment in high-noise scene

Abstract

The invention discloses an intelligent communication system and equipment in a high-noise scene, and relates to the technical field of communication, the system comprises a communication host, a Bluetooth pickup, a head-mounted loudspeaker assembly and a command and dispatch platform; the Bluetooth pickup is connected with a positive pressure type air respirator of a wearer and used for collecting voice signals and transmitting the voice signals to the communication host through Bluetooth, and the communication host transmits the signals to the commanding and dispatching platform after the voice signals are processed through a noise reduction algorithm. According to the intelligent communication system and equipment in the high-noise scene, deep learning and adaptive gain control technologies are utilized, the communication quality, voice definition and system response speed are remarkably improved in the high-noise scene, meanwhile, the defects of a traditional technology in a complex noise environment are overcome, and the intelligent communication system and equipment in the high-noise scene have a wide application prospect. The requirements that efficient communication and equipment control still need to be carried out when the two hands of a fireman are occupied are met, powerful technical support is provided for high-risk tasks such as fire rescue, and high efficiency and safety of task execution are ensured.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an intelligent communication system and equipment in a high noise scenario. Background Art

[0002] In many high-noise scenarios such as firefighting and rescue, traditional communication equipment and systems are difficult to meet actual needs and have many problems. Firefighters usually need to wear positive pressure air respirators when performing firefighting and rescue tasks, which will produce a series of factors that affect communication. On the one hand, the noise in the fire scene is complex and loud, often covering up human voices, seriously interfering with the quality of communication, and making it difficult for firefighters to clearly transmit information to each other and to the command center. On the other hand, the breathing noise caused by normal breathing of the positive pressure air respirator also has an adverse effect on communication. Existing communication technology means have obvious defects in dealing with these problems. For example, dual-microphone noise reduction technology can resist the interference of environmental noise around mobile phones to a certain extent through physical methods and simple algorithms, but in the face of the complex and high-intensity noise environment at the fire scene, its noise reduction effect is not good, and it cannot effectively identify sound signals. It is difficult to meet the needs of firefighting and rescue scenarios, and it cannot change the design defects of the walkie-talkie itself, such as channel occupancy. Although bone conduction technology has certain advantages in restoring sound in noisy environments, it converts sound into mechanical vibrations and transmits sound waves through the skull, etc., which is divided into bone conduction speaker technology and bone conduction microphone technology. However, bone conduction speaker technology and bone conduction microphone technology cannot be used at the same time, and cannot solve the problem of firefighters hearing and speaking without noise at the same time. Moreover, when firefighters wear existing firefighting equipment, the bone conduction device needs to be close to the skull, which is extremely uncomfortable to wear. As an accessory of the walkie-talkie, it cannot change the communication difficulties caused by the design principle of the walkie-talkie itself. In summary, the existing communication technology cannot effectively guarantee the communication quality, meet the operation requirements of firefighters when both hands are occupied, and achieve efficient equipment control and information interaction in high-noise fire rescue scenarios. A new intelligent communication system and equipment are urgently needed to solve these problems. Summary of the invention

[0003] In view of the shortcomings of the existing technology, the present invention provides an intelligent communication system and equipment for high-noise scenarios, which solves the problem that the existing communication technology is unable to effectively guarantee the communication quality, meet the operational requirements of firefighters when both hands are occupied, and realize efficient equipment control and information interaction in high-noise fire rescue scenarios.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent communication system and equipment in a high noise scene, the system includes a communication host, a Bluetooth microphone, a head-mounted speaker assembly and a command and dispatch platform, wherein the Bluetooth microphone is connected to the wearer's positive pressure air respirator, used to collect voice signals and transmit them to the communication host via Bluetooth, and after being processed by a noise reduction algorithm, the communication host transmits the signal to the command and dispatch platform; the command and dispatch platform is used to receive signals and exchange information; The communication host has a built-in Android operating system, a noise reduction chip and a host chip, which are used to receive audio signals, perform noise reduction processing and signal distribution, and has a built-in self-developed terminal APP, which supports automatic login and group calls according to preset groups, and voice interaction can be achieved without additional operations; The Bluetooth microphone is arranged in the air-exhalation mask of the positive pressure air respirator, and voice collection is performed through a microphone installed in the mask of the positive pressure air respirator, which can effectively reduce the influence of breathing noise on communication quality, and is used to pick up sound and transmit it to the communication host via Bluetooth; The head-mounted speaker assembly includes an integrated microphone and speaker, which can transmit voice signals through cables, and emit sound after noise reduction processing under the processing of the communication host, thereby providing a clear communication effect; in non-firefighting and rescue tasks, the microphone of the head-mounted speaker assembly is used to pick up sound, and transmit the audio signal to the noise reduction chip of the communication host through cables for algorithm noise reduction processing, thereby ensuring the clarity and stability of the voice signal; Among them, the communication host supports voice command control, allowing firefighters to control the equipment through voice commands, including single-point calling the command center and single-point calling personnel; the communication host also transmits information through the 4G network, supports cross-regional real-time communication, and can exchange information between different rescue sites; in addition, the communication host can also receive sound information from other devices, and distribute it to the head-mounted speaker component through the host chip via cables for external playback.

[0005] Preferably, the noise reduction chip loads different artificial intelligence noise reduction algorithms according to the set noise scenes, so as to eliminate the impact of noise on communication quality in fire rescue scenes, so as to improve the speech recognition rate in complex environments; the artificial intelligence noise reduction algorithm trains different noise samples based on a deep learning model, so as to adapt to noise interference in different scenes and automatically optimize the clarity of the speech signal.

[0006] Preferably, the process of the noise reduction chip loading different artificial intelligence noise reduction algorithms according to the set noise scene includes: The noise reduction chip loads a deep learning-based explosion noise model according to the detected high-intensity background noise characteristics. The model analyzes the instantaneous frequency characteristics of high-frequency explosion sounds, uses time-frequency analysis and dynamic filtering algorithms, effectively removes the interference of such explosion noise on voice signals, and optimizes the clarity of voice signals. The noise reduction chip loads a low-frequency noise model based on deep learning according to the detected low-frequency noise characteristics. The model identifies low-frequency noise through spectrum analysis and deep convolutional neural network, and uses low-pass filtering and spectrum subtraction algorithms to remove noise, while enhancing the clarity of the voice signal through adaptive gain control; The noise reduction chip loads a high-frequency noise model based on deep learning according to the detected intermittent high-frequency noise characteristics. The model identifies high-frequency noise through periodic noise detection and time-frequency analysis methods, removes high-frequency noise through bandpass filtering and noise suppression algorithms, and enhances the clarity of voice signals.

[0007] The noise reduction chip loads a deep learning-based breathing noise model according to the characteristics of breathing noise generated when wearing a positive pressure air ventilator. The model recognizes the periodic characteristics of breathing sounds, uses spectrum separation and adaptive algorithms to remove breathing noise, and optimizes the clarity of voice signals.

[0008] Preferably, the noise reduction chip loads an explosion noise model based on deep learning according to the detected high-intensity background noise characteristics, and the model converts the audio signal by using a time-frequency analysis method to generate a time-frequency diagram by analyzing the instantaneous frequency characteristics of the high-frequency explosion sound; the deep learning explosion noise model is trained by a convolutional neural network CNN or a recursive neural network RNN, and can identify the frequency, amplitude and time characteristics of the explosion noise, and dynamically adjust the parameters of the filtering algorithm according to the recognition results, thereby effectively removing the interference of the explosion noise on the voice signal; the noise reduction chip further includes a dynamic filtering algorithm, which performs adaptive filtering based on the characteristics of the explosion noise, optimizes the clarity of the voice signal, and ensures the high-quality transmission of the remaining voice part; the time-frequency analysis method generates a time-frequency diagram by short-time Fourier transform STFT or continuous wavelet transform CWT, monitors the intensity change of the explosion noise in real time, and adjusts the noise reduction strategy according to the instantaneous frequency of the explosion noise, thereby ensuring the accuracy of the noise removal effect; the noise reduction chip also includes a voice enhancement module, which, after removing the explosion noise, enhances the audibility and clarity of the remaining voice signal by adaptive gain control technology, thereby improving the quality of the voice signal and ensuring effective communication.

[0009] Preferably, the noise reduction chip loads a low-frequency noise model based on deep learning according to the detected low-frequency noise characteristics. The model identifies the characteristics of low-frequency noise through a spectrum analysis method, and uses a deep convolutional neural network (CNN) for training to accurately identify the difference between low-frequency noise and speech signals; the spectrum analysis method generates a spectrum diagram of the audio signal through the fast Fourier transform (FFT) technology to identify the frequency range and amplitude of the low-frequency noise signal; the low-frequency noise model processes the low-frequency noise in the audio signal in real time according to the low-frequency noise characteristics in the training data set, removes the low-frequency noise through a low-pass filtering algorithm, and retains the high-frequency part of the speech signal; the noise reduction chip further removes the low-frequency noise through a spectrum subtraction algorithm, which subtracts the spectrum part of the low-frequency noise in the time-frequency domain, thereby effectively eliminating the interference of low-frequency noise on the speech; the noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to the change in signal strength, thereby enhancing the clarity of the speech signal and ensuring the intelligibility of the speech signal in a complex environment.

[0010] Preferably, the noise reduction chip loads a high-frequency noise model based on deep learning according to the detected intermittent high-frequency noise characteristics. The model identifies high-frequency noise through periodic noise detection and time-frequency analysis methods, and removes high-frequency noise through bandpass filtering and noise suppression algorithms; the periodic noise detection and time-frequency analysis method performs time-frequency conversion on the audio signal through short-time Fourier transform STFT or continuous wavelet transform CWT to generate a time-frequency graph, and analyzes the periodic characteristics and frequency fluctuations of the high-frequency noise; the deep learning high-frequency noise model is trained by a deep convolutional neural network CNN or a recursive neural network RNN, and can extract the characteristics of high-frequency noise from the time-frequency graph and distinguish it from the speech signal, and identify and suppress intermittent high-frequency noise in real time; the bandpass filter filters a specific frequency range of the audio signal to remove the high-frequency noise part in the audio signal while retaining the effective frequency in the speech signal; the noise suppression algorithm removes the noise signal overlapping with the speech through spectral subtraction or frequency domain filtering methods, and optimizes the clarity of the remaining speech signal through speech enhancement technology; The noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to changes in signal strength, thereby enhancing the clarity of the voice signal and ensuring the intelligibility of the voice signal in a complex noise environment.

[0011] Preferably, the noise reduction chip loads a deep learning-based breathing noise model according to the characteristics of breathing noise generated when wearing a positive pressure air ventilator. The model recognizes the periodic characteristics of breathing sounds, uses spectrum separation and adaptive algorithms to remove breathing noise, and optimizes the clarity of voice signals. The deep learning breathing noise model is trained by a convolutional neural network (CNN) or a long short-term memory network (LSTM), and can identify and separate the breathing noise part in the audio signal, and distinguish it from the frequency components of the speech signal; the spectrum separation method generates a time-frequency diagram of the audio signal through time-frequency analysis technology, such as short-time Fourier transform (STFT) or wavelet transform (CWT), and identifies and removes the spectrum part of the breathing noise through spectrum analysis; the adaptive algorithm dynamically adjusts the intensity of noise removal according to the change of the signal through Kalman filtering and minimum mean square error (MMSE), and optimizes the effective components in the speech signal; The noise reduction chip further includes a speech enhancement module, which improves the quality of speech signals through speech enhancement technology, reduces the influence of residual breathing noise, and ensures the clarity and comprehensibility of speech signals; The noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to changes in signal strength to ensure clear transmission of voice signals in complex environments.

[0012] Preferably, the noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to changes in signal strength to enhance the clarity of the voice signal and ensure the intelligibility of the voice signal in a complex noise environment; the adaptive gain control module calculates the ratio of the voice signal to the noise signal by real-time monitoring the amplitude and noise intensity of the audio signal, and adjusts the gain value accordingly to adapt the volume of the voice signal to changes in the ambient noise intensity; the gain calculation method includes adjusting the gain value according to the dynamic change of the signal strength, if the noise intensity is large, increasing the gain value to enhance the clarity of the voice signal; if the noise intensity is small, reducing the gain value to avoid distortion of the voice signal; the gain adjustment process adopts a smoothing control algorithm to avoid distortion due to sudden gain changes and ensure the continuity and naturalness of the voice signal; the gain control module is combined with a voice enhancement algorithm to optimize the clarity of the signal and enhance the audibility and intelligibility of the voice signal in a complex noise environment.

[0013] Preferably, the voice command control system includes a voice recognition module and a control module, the voice recognition module can recognize and process voice commands, and the control module performs corresponding control operations according to the recognition results.

[0014] An intelligent communication device for high noise scenarios, comprising an intelligent communication system for high noise scenarios, and a positive pressure air respirator and a command and dispatch platform used in conjunction with the intelligent communication system, wherein: Positive pressure air respirator, which provides breathing support for firefighters, and has a Bluetooth microphone installed on its air respiration mask; The command and dispatch platform is deployed on a tablet or PC and connected to the communication host through a 4G network. It performs group setting, positioning display, single call, group call, temporary group, and broadcast according to the information collected in real time, so as to efficiently coordinate the communication between firefighters and the command center.

[0015] The present invention provides an intelligent communication system and device in a high noise scene, which has the following beneficial effects: This intelligent communication system and equipment in high-noise scenarios uses deep learning and adaptive gain control technology to significantly improve communication quality, voice clarity and system response speed in high-noise scenarios. At the same time, it solves the shortcomings of traditional technologies in complex noise environments and meets the needs of firefighters for efficient communication and equipment control when their hands are occupied. It provides strong technical support for high-risk tasks such as firefighting and rescue, and ensures the efficiency and safety of task execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a framework of an intelligent communication system in a high noise scenario according to the present invention; Figure 2 It is a schematic diagram of the process of noise scene detection of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] See also Figure 1 and Figure 2 The present invention provides a technical solution: an intelligent communication system and equipment in a high noise scene, the system includes a communication host, a Bluetooth microphone, a head-mounted speaker assembly and a command and dispatch platform, wherein the Bluetooth microphone is connected to the wearer's positive pressure air respirator, and is used to collect voice signals and transmit them to the communication host via Bluetooth. After being processed by a noise reduction algorithm, the communication host transmits the signal to the command and dispatch platform; the command and dispatch platform is used to receive signals and exchange information; The communication host has built-in Android operating system, noise reduction chip and host chip, which are used to receive audio signals, perform noise reduction processing and signal distribution. It also has built-in self-developed terminal APP, which supports automatic login and group calls according to preset groups, and voice interaction can be achieved without additional operations. The Bluetooth microphone is set in the air-exhalation mask of the positive pressure air respirator. Voice collection is carried out through the microphone installed in the mask of the positive pressure air respirator, which can effectively reduce the impact of breathing noise on communication quality. It is used to pick up sound and transmit it to the communication host via Bluetooth. The headset speaker assembly includes an integrated microphone and speaker, which can transmit voice signals through cables and emit sound after noise reduction processing under the processing of the communication host, thereby providing clear communication effects; in non-firefighting and rescue missions, the microphone of the headset speaker assembly is used to pick up the sound, and transmit the audio signal through the cable to the noise reduction chip of the communication host for algorithm noise reduction processing, thereby ensuring the clarity and stability of the voice signal; Among them, the communication host supports voice command control, allowing firefighters to control the equipment through voice commands, including single-point calls to the command center and single-point calls to personnel; the communication host also transmits information through the 4G network, supports cross-regional real-time communication, and can exchange information between different rescue sites; in addition, the communication host can also receive sound information from other devices, and distribute it to the head-mounted speaker component through the host chip via cables for external playback.

[0019] The noise reduction chip loads different artificial intelligence noise reduction algorithms according to the set noise scenarios, which is used to eliminate the impact of noise on communication quality in fire rescue scenarios, so as to improve the voice recognition rate in complex environments; the artificial intelligence noise reduction algorithm trains different noise samples based on the deep learning model to adapt to noise interference in different scenarios and automatically optimize the clarity of the voice signal.

[0020] The process of the noise reduction chip loading different artificial intelligence noise reduction algorithms according to the set noise scene includes: The noise reduction chip loads a deep learning-based explosion noise model according to the detected high-intensity background noise characteristics. The model converts the audio signal using a time-frequency analysis method by analyzing the instantaneous frequency characteristics of the high-frequency explosion sound to generate a time-frequency diagram. The deep learning explosion noise model is trained through a convolutional neural network (CNN), which can identify the frequency, amplitude and time characteristics of the explosion noise, and dynamically adjust the parameters of the filtering algorithm according to the recognition results, thereby effectively removing the interference of the explosion noise on the voice signal. The noise reduction chip further includes a dynamic filtering algorithm, which performs adaptive filtering based on the characteristics of the explosion noise and optimizes the clarity of the voice signal to ensure high-quality transmission of the remaining voice part. The time-frequency analysis method generates a time-frequency diagram through a short-time Fourier transform (STFT), monitors the intensity changes of the explosion noise in real time, and adjusts the noise reduction strategy according to the instantaneous frequency of the explosion noise, thereby ensuring the accuracy of the noise removal effect. The noise reduction chip also includes a voice enhancement module, which, after removing the explosion noise, enhances the audibility and clarity of the remaining voice signal through an adaptive gain control technology, thereby improving the quality of the voice signal and ensuring effective communication. It should be further explained that, in the specific implementation process, first, the noise reduction chip collects environmental audio signals in real time through the built-in microphone array; to ensure that the characteristics of high-intensity background noise are accurately captured, the noise reduction chip monitors the instantaneous frequency and amplitude of the audio signal; by using the short-time Fourier transform STFT, the noise reduction chip can obtain detailed information of the signal from the time-frequency domain; the noise reduction chip performs time-frequency conversion on the collected audio signal to generate a time-frequency graph; through the time-frequency graph, the system can identify the instantaneous frequency characteristics of the explosion noise; for example, the explosion sound usually has an instantaneous high-frequency energy release, which is manifested as a sudden high-amplitude, high-frequency area in the time-frequency graph; the noise reduction chip loads a trained "explosion noise" model based on deep learning according to the high-frequency explosion noise characteristics extracted from the time-frequency graph; the model uses the convolutional neural network CNN in the deep neural network to identify and distinguish the difference between the explosion sound and human speech; The model uses a large amount of explosion noise sample data during the training process, including different types of explosions and impact sounds. The model learns how to separate the characteristics of explosion noise from normal voice signals. The training data includes explosion noise and voice data in various environments to ensure that the model can adapt to complex noise scenarios. After training, the deep learning model can identify the instantaneous frequency, amplitude, amplitude envelope and temporal burst characteristics of explosion noise. The model analyzes these characteristics in the audio signal and accurately separates the explosion noise from the voice signal. Through the output of the explosion noise model, the noise reduction chip can effectively remove the explosion noise through a dynamic filtering algorithm during real-time processing. During the noise reduction process, the noise reduction chip uses the Kalman filter in the dynamic adaptive filtering algorithm to adjust the filter parameters to remove the impact of explosion noise in real time; these filtering algorithms adaptively adjust the frequency and amplitude of the processed signal according to the explosion noise characteristics output by the model, thereby accurately removing noise; The filter first subtracts the corresponding noise from the audio signal according to the frequency range and amplitude of the noise; in actual applications, the filtering algorithm will adjust in real time to cope with explosion noise of different intensities and frequencies, ensuring that only the speech component is retained in the audio signal; After the explosion noise is removed, the noise reduction chip continues to optimize the remaining voice signal to improve the clarity and intelligibility of the voice; the de-noised voice signal is further processed using the linear predictive coding (LPC) in the voice enhancement technology; this technology can enhance the clarity and naturalness of the voice signal and reduce the interference of residual noise on the voice; according to other background noises in the environment, such as wind and machine operation, the noise reduction chip uses an adaptive gain control algorithm to automatically adjust the gain of the voice signal to achieve the best balance of clarity in a noisy environment; after dynamic filtering, noise removal and signal enhancement, the quality of the voice signal is guaranteed and can maintain a high level of clarity. Even in the background of high-intensity explosion noise, the voice can still be accurately recognized and transmitted; In actual use, the noise reduction chip will continuously monitor the noise environment and make adaptive adjustments according to environmental changes; if the frequency, intensity or other characteristics of the explosion noise change, the noise reduction chip will automatically update the parameters of the filtering algorithm and deep learning model to ensure that the noise reduction effect is always maintained in the optimal state; the noise reduction chip combines the feedback mechanism to continuously optimize the working parameters of the filter by monitoring the changes of noise and voice signals in real time; even if the explosion noise changes in the spectrum, the system can make corresponding adjustments in time to maintain the clarity of the voice; Finally, after the above processing, the noise reduction chip will transmit the optimized voice signal through the communication host to other devices, such as the command and dispatch platform; in this process, the voice signal will be sent via Bluetooth or other transmission technologies and will not be affected by high-intensity background noise, thereby ensuring real-time and clear voice communication.

[0021] The noise reduction chip loads a low-frequency noise model based on deep learning according to the detected low-frequency noise characteristics. The model identifies the characteristics of low-frequency noise through a spectrum analysis method and uses a deep convolutional neural network (CNN) for training to accurately identify the difference between low-frequency noise and speech signals. The spectrum analysis method generates a spectrum diagram of the audio signal through the fast Fourier transform (FFT) technology to identify the frequency range and amplitude of the low-frequency noise signal. The low-frequency noise model processes the low-frequency noise in the audio signal in real time according to the low-frequency noise characteristics in the training data set, removes the low-frequency noise through a low-pass filtering algorithm, and retains the high-frequency part of the speech signal. The noise reduction chip further removes the low-frequency noise through a spectrum subtraction algorithm, which subtracts the spectrum part of the low-frequency noise in the time-frequency domain, thereby effectively eliminating the interference of low-frequency noise on speech. The noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to the change in signal strength, thereby enhancing the clarity of the speech signal and ensuring the intelligibility of the speech signal in complex environments. It should be further explained that in the specific implementation process, the noise reduction chip first collects audio signals in real time through sensors such as microphone arrays. In order to detect low-frequency noise, such as fan noise and mechanical equipment operation noise, the system uses fast Fourier transform FFT to perform spectrum analysis on the audio signal. The audio signal is converted into a spectrum through FFT, and the noise reduction chip can identify the frequency characteristics of the low-frequency noise signal. Low-frequency noise is usually concentrated in the frequency band of 20Hz to 500Hz. These noises show a relatively constant frequency pattern in the spectrum, and the noise reduction chip identifies these patterns through spectrum analysis. The system extracts the frequency, amplitude and time domain characteristics in the audio signal to determine which frequency components belong to low-frequency noise and distinguish them from the voice signal. The noise reduction chip monitors the changes of these low-frequency noises in real time and determines whether they need to be removed.

[0022] According to the low-frequency noise features extracted from the spectrum, the noise reduction chip loads a low-frequency noise recognition model based on deep learning, which is trained through a deep convolutional neural network (CNN) to specifically identify and remove low-frequency noise. The low-frequency noise model learns the spectral features of low-frequency noise by using a large amount of training data with low-frequency noise, including fan noise, air conditioning noise, and engine running sound. During the training process, CNN extracts the time-frequency features of low-frequency noise and optimizes the model parameters through back propagation to efficiently identify and remove these noises in a real-time environment. The deep learning model analyzes the spectral data of the audio signal and determines which parts belong to low-frequency noise. The model can accurately distinguish between low-frequency noise and voice signals, so as to perform targeted processing.

[0023] In order to effectively remove low-frequency noise, the noise reduction chip combines low-pass filtering with a spectrum subtraction algorithm. The low-pass filter allows signals below a certain set frequency to pass through while weakening high-frequency signals. The noise reduction chip uses a low-pass filter to filter out the low-frequency noise part and remove the noise components with lower frequencies. This step can significantly reduce the impact of low-frequency noise on the voice signal. Spectral subtraction is a noise removal method based on time-frequency analysis. After the noise reduction chip performs time-frequency conversion on the audio signal, it uses a spectrum subtraction algorithm to subtract the low-frequency noise components and eliminate the low-frequency part of the noise signal. In this way, the system can remove low-frequency noise while retaining the key components of the voice signal.

[0024] After removing low-frequency noise, the noise reduction chip further enhances the remaining voice signal to improve the clarity and audibility of the voice. After removing low-frequency noise, the noise reduction chip uses linear predictive coding (LPC) to optimize the voice signal. This technology can improve the quality of speech and enhance the voice frequency components in the audio, making the speech clearer and easier to understand. According to the intensity of the audio signal and the changes in ambient noise, the noise reduction chip automatically adjusts the gain to ensure that the voice signal maintains the appropriate volume and clarity in different noise environments. Adaptive gain control can adjust the gain factor based on the results of real-time analysis, so that the voice signal is enhanced to the greatest extent after the low-frequency noise is removed.

[0025] In actual use, the noise reduction chip can continuously monitor changes in the audio environment and adjust the processing strategy according to the real-time changes in the noise environment. The sensors built into the noise reduction chip continuously collect environmental noise information and adjust the parameters of the filter and noise reduction algorithm based on the low-frequency noise characteristics analyzed in real time. The system will adaptively adjust according to changes in the environment to ensure that the voice signal remains clear and natural under continuous noise changes. Finally, the noise reduction chip transmits the optimized voice signal to the communication host or command and dispatch platform to ensure that the voice signal is clear and audible without interference from low-frequency noise.

[0026] This technical process describes in detail how the noise reduction chip loads the deep learning low-frequency noise model based on the detected low-frequency noise characteristics, removes low-frequency noise through spectral analysis, low-pass filtering, spectral subtraction and other algorithms, and enhances the clarity of voice signals through adaptive gain control. This process ensures high-quality transmission of voice signals in complex noise environments and effectively solves the problem of low-frequency noise interfering with voice communications.

[0027] The noise reduction chip loads a high-frequency noise model based on deep learning according to the detected intermittent high-frequency noise characteristics. The model identifies high-frequency noise through periodic noise detection and time-frequency analysis methods, and removes high-frequency noise through bandpass filtering and noise suppression algorithms. The periodic noise detection and time-frequency analysis methods perform time-frequency conversion on the audio signal through short-time Fourier transform STFT or continuous wavelet transform CWT to generate a time-frequency graph and analyze the periodic characteristics and frequency fluctuations of high-frequency noise. The deep learning high-frequency noise model is trained through deep convolutional neural network CNN or recurrent neural network RNN, which can extract the characteristics of high-frequency noise from the time-frequency graph and distinguish it from the speech signal, and identify and suppress intermittent high-frequency noise in real time. The bandpass filter filters the specific frequency range of the audio signal to remove the high-frequency noise part in the audio signal while retaining the effective frequency in the speech signal. The noise suppression algorithm removes the noise signal overlapping with the speech through spectral subtraction or frequency domain filtering methods, and optimizes the clarity of the remaining speech signal through speech enhancement technology. The noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to changes in signal strength, thereby enhancing the clarity of the voice signal and ensuring the intelligibility of the voice signal in a complex noisy environment; It should be further explained that, in the specific implementation process, first, the noise reduction chip collects the environmental audio signal in real time through the microphone array or other sensors; in this process, the system pays special attention to the high-frequency part of the audio signal, especially the intermittent high-frequency noise, such as alarm sound, electronic equipment noise, and equipment startup; the noise reduction chip uses short-time Fourier transform STFT or continuous wavelet transform CWT to transform the audio signal to generate a time-frequency diagram; through the time-frequency diagram, the system can clearly identify the instantaneous frequency changes and characteristics of the high-frequency noise, for example, intermittent high-frequency noise usually has a periodic fluctuation pattern; through the periodic noise detection algorithm, the noise reduction chip can identify the periodic high-frequency noise in the audio signal; these noises usually repeat at a fixed frequency or interval, and the model detects and confirms these noises through quantitative analysis of the frequency; After detecting high-frequency noise, the noise reduction chip loads a high-frequency noise recognition model trained by deep learning. The model can accurately identify intermittent high-frequency noise and distinguish it from voice signals. The model is trained using a data set containing a large number of high-frequency noise samples. During the training process, deep learning models, such as convolutional neural networks (CNN) or recurrent neural networks (RNN), learn the periodic characteristics of noise and can accurately identify various types of intermittent high-frequency noise, such as alarm sounds and mechanical equipment operation sounds. After training, the deep learning model can extract the spectral characteristics of high-frequency noise from the audio signal, analyze the time intervals and frequency changes of these noises, thereby separating voice and noise, and accurately identifying the interference part of the noise. In order to remove intermittent high-frequency noise, the noise reduction chip combines bandpass filtering with noise suppression algorithms to ensure that the quality of the voice signal is not affected by noise; the bandpass filter can filter the noise part of the audio signal according to the frequency range of high-frequency noise, such as the frequency band above 3000Hz; by setting the appropriate bandpass frequency band, the filter allows the high-frequency part of the voice signal to pass through, while suppressing the frequency band of high-frequency noise; the noise suppression algorithm further removes high-frequency noise by calculating the noise component of the audio signal and distinguishing it from the voice component; the algorithm identifies and subtracts the energy of the noise signal through spectral analysis to retain the effective information in the voice signal; this process helps to eliminate distortion caused by high-frequency noise; After high-frequency noise is removed, the noise reduction chip further enhances the remaining speech signal to improve the clarity and audibility of the speech. The noise reduction chip improves the quality of the speech signal by applying speech enhancement algorithms, such as linear predictive coding (LPC) or speech reconstruction technology. These technologies can enhance the audibility of the speech in a noisy environment while preserving the original characteristics of the speech. Adaptive gain control: The adaptive gain control module automatically adjusts the gain according to the strength of the audio signal. By dynamically adjusting the gain coefficient, the system can ensure that the speech signal remains clear and understandable under different noise conditions. The noise reduction chip monitors the changes in the audio environment in real time and automatically adjusts the processing strategy to ensure that the system always provides the best noise removal and voice enhancement effects under different noise conditions; Environmental monitoring and adaptive adjustment: The noise reduction chip continuously collects audio signals and dynamically adjusts the frequency range of the bandpass filter and the parameters of the noise suppression algorithm according to the changes in environmental noise; the system can identify high-frequency noise changes in audio signals in real time and quickly adapt to new noise patterns; finally, the voice signal after denoising and enhancement processing will be transmitted to other devices such as the command and dispatch platform or receiving device through the communication host to ensure the clarity and comprehensibility of the voice signal in a high-frequency noise environment; The above technical process describes in detail how to use the noise reduction chip to load the deep learning high-frequency noise model according to the intermittent high-frequency noise characteristics to remove noise and enhance speech; through periodic noise detection, time-frequency analysis, bandpass filtering, noise suppression and adaptive gain control and other technologies, the noise reduction chip can effectively remove high-frequency noise and optimize the clarity and audibility of the voice signal, thereby ensuring the quality of voice communication in a high-noise environment.

[0028] The noise reduction chip loads a deep learning-based breathing noise model based on the characteristics of breathing noise generated when wearing a positive pressure air ventilator. The model recognizes the periodic characteristics of breathing sounds, uses spectrum separation and adaptive algorithms to remove breathing noise, and optimizes the clarity of voice signals. The deep learning breathing noise model is trained through convolutional neural network CNN or long short-term memory network LSTM, which can identify and separate the breathing noise part in the audio signal and distinguish it from the frequency components of the speech signal; the spectrum separation method generates the time-frequency diagram of the audio signal through time-frequency analysis technology, such as short-time Fourier transform STFT or wavelet transform CWT, and identifies and removes the spectrum part of the breathing noise through spectrum analysis; the adaptive algorithm dynamically adjusts the intensity of noise removal according to the change of the signal through Kalman filtering and minimum mean square error MMSE, and optimizes the effective components in the speech signal; the noise reduction chip further includes a speech enhancement module, which improves the quality of the speech signal through speech enhancement technology, reduces the influence of residual breathing noise, and ensures the clarity and comprehensibility of the speech signal; the noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to the change of signal strength to ensure the clear transmission of the speech signal in a complex environment; It should be further explained that, in the specific implementation process, the noise reduction chip first collects audio signals in real time through a microphone array or other sensors; this process pays special attention to the breathing noise generated when wearing a positive pressure air ventilator; the positive pressure air ventilator will generate continuous breathing noise during use, and these noises have periodicity and certain frequency characteristics, usually in the low frequency range; the noise reduction chip uses time-frequency analysis methods, such as short-time Fourier transform STFT or continuous wavelet transform CWT, to perform spectral analysis on the audio signal; this step can clearly detect the frequency characteristics of breathing noise, for example, breathing noise is usually concentrated in a lower frequency range: 20Hz to 500Hz, and presents a periodic fluctuation pattern; the noise reduction chip uses a periodic feature extraction algorithm to identify the periodic features related to breathing noise in the audio signal, which is specifically manifested as a noise pattern that repeats periodically; this method can accurately distinguish the difference between breathing noise and speech signals; After the breathing noise is detected, the noise reduction chip loads a breathing noise model trained by deep learning. The model can identify and separate breathing noise from valid speech signals in audio signals by learning the periodic characteristics of breathing noise. Deep learning models, such as convolutional neural networks (CNN) or long short-term memory networks (LSTM), use a large amount of breathing noise sample data during training, such as various breathing patterns of people wearing positive pressure air respirators, and learn the periodicity, frequency and waveform characteristics of breathing sounds through the model. The model can effectively identify and separate breathing noise, especially in complex background noise environments. Based on the periodic noise characteristics in the training data, the deep learning model can extract patterns related to breathing noise from audio signals and separate them by determining the difference between them and speech signals. Through the output of the deep learning model, the noise reduction chip can remove the identified breathing noise from the audio signal and optimize the voice signal; the noise reduction chip uses spectrum separation technology to distinguish the breathing noise frequency part in the audio signal from the frequency part of the voice signal; the spectrum separation method can effectively isolate and remove the breathing noise component, thereby reducing the interference of breathing noise on the voice signal; the noise reduction chip further adjusts the parameters of the filter through adaptive denoising algorithms, such as Kalman filtering and minimum mean square error filtering, to dynamically remove breathing noise; the algorithm adjusts the degree of noise removal in real time according to the changes in the noise intensity in the audio signal to ensure that the removal of breathing noise does not affect the clarity of the voice signal; After the breathing noise is removed, the noise reduction chip further optimizes the remaining voice signal to improve the clarity and audibility of the voice. The noise reduction chip uses a voice enhancement algorithm, such as linear predictive coding (LPC) or a voice reconstruction algorithm, to process the denoised voice signal. Voice enhancement technology can improve the quality of voice and enhance its clarity, making the voice signal easier to understand in a complex noisy environment. In order to ensure the stability and clarity of the voice signal, the noise reduction chip also uses adaptive gain control technology. This technology dynamically adjusts the gain coefficient according to the signal strength and changes in ambient noise, thereby optimizing the volume and clarity of the voice signal, ensuring that the voice can be clearly conveyed in any environment. The noise reduction chip monitors the changes in the audio environment in real time and automatically adjusts the processing strategy to ensure the best voice clarity under different noise conditions. Noise adaptation and dynamic adjustment: The system continuously monitors the characteristics of breathing noise and changes in background noise, and adjusts the parameters of the filter and denoising algorithm in real time. No matter how the environment changes, the noise reduction chip can adapt to and optimize the noise removal effect according to the new noise pattern. Finally, the voice signal after denoising and enhancement processing is transmitted to the receiving device through the communication host or command and dispatch platform to ensure the clarity and comprehensibility of the voice signal.

[0029] The noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to changes in signal strength to enhance the clarity of the voice signal and ensure the intelligibility of the voice signal in a complex noise environment; the adaptive gain control module calculates the ratio of the voice signal to the noise signal by real-time monitoring the amplitude and noise intensity of the audio signal, and adjusts the gain value accordingly to adapt the volume of the voice signal to changes in the ambient noise intensity; the gain calculation method includes adjusting the gain value according to the dynamic change of the signal strength, if the noise intensity is large, the gain value is increased to enhance the clarity of the voice signal; if the noise intensity is small, the gain value is reduced to avoid distortion of the voice signal; the gain adjustment process adopts a smoothing control algorithm to avoid distortion due to sudden gain changes and ensure the continuity and naturalness of the voice signal; the gain control module combines the voice enhancement algorithm to optimize the clarity of the signal and enhance the audibility and intelligibility of the voice signal in a complex noise environment.

[0030] The voice command control system includes a voice recognition module and a control module. The voice recognition module can recognize and process voice commands, and the control module performs corresponding control operations according to the recognition results. The noise reduction chip first collects audio signals in real time through a microphone array. These signals will vary depending on changes in environmental noise, so it is necessary to monitor the signal strength and quality in real time.

[0031] It should be further explained that in the specific implementation process, the noise reduction chip monitors the amplitude of the audio signal in real time and calculates the average power or amplitude of the signal to understand the strength of the voice signal and background noise. The system evaluates the power of the voice signal and compares it with the background noise to ensure that the effective voice signal is not masked by the noise. In order to ensure the accuracy of signal enhancement, the noise reduction chip also analyzes the noise characteristics in the environment through a noise recognition algorithm. Especially in complex environments, the system will identify the noise source in real time and calculate its impact on the signal to determine the amplitude that needs gain control.

[0032] According to the real-time detected signal strength and noise level, the adaptive gain control module of the noise reduction chip will adjust the gain to enhance the clarity of the voice signal and maintain the naturalness of the voice. The gain control module calculates the appropriate gain value based on the ratio of signal strength to noise strength. If the noise in the environment is large and the voice signal is weak, the gain module will increase the gain value to enhance the strength of the voice signal; if the background noise is small or the voice signal itself is strong, the gain value will be reduced accordingly to avoid distortion or over-amplification of the voice signal.

[0033] The gain control module optimizes the gain according to the real-time changes in signal strength through the dynamic adjustment algorithm MMSE algorithm and Kalman filter algorithm. This ensures that the voice signal remains clear and discernible in different noise environments. In order to avoid sudden changes or distortion during the gain adjustment process, the system uses smoothing control technology to make the gain change smoother and more gradual. This helps to maintain the naturalness of the voice and avoid abrupt changes in voice quality.

[0034] The adaptive gain control module not only adjusts the gain, but also further optimizes the voice signal to ensure high-quality transmission of the voice signal in different noise environments. After the gain is adjusted, the system uses linear predictive coding LPC to further enhance the clarity and audibility of the voice. These algorithms can remove unnecessary frequency bands in the voice signal and enhance the frequency response of the voice, thereby improving the quality of the voice. In a high-noise environment, the noise reduction chip continuously adjusts the gain through the adaptive gain control module, so that the voice signal always remains clear under complex noise conditions. The system automatically adapts to changes in environmental noise to ensure that the voice signal can be accurately transmitted in noise interference.

[0035] Finally, after gain adjustment and voice enhancement, the optimized voice signal is transmitted through the communication host or other devices to ensure the clarity and intelligibility of the voice signal. The optimized voice signal will be transmitted to the receiving device without noise interference, ensuring the efficiency and accuracy of communication.

[0036] This technical process describes in detail the working principle and technical process of the adaptive gain control module in the noise reduction chip. By monitoring the signal strength and background noise in real time, the gain is automatically calculated and adjusted to enhance the clarity of the voice signal. In a complex noise environment, the module can dynamically adapt to environmental changes, maintain the naturalness and audibility of the voice signal, and ensure efficient and accurate voice transmission during the communication process.

[0037] It should be further explained that in fire rescue missions, firefighters usually encounter the following two major problems when wearing positive pressure air respirators: first, the environmental noise at the fire scene will seriously affect the quality of communication, and second, the breathing noise generated by wearing positive pressure air respirators will also affect the clarity of communication. In addition, firefighters' hands are usually unable to freely operate communication equipment when performing rescue missions, resulting in operational inconvenience.

[0038] An intelligent communication system is provided, which includes a communication host, a Bluetooth microphone, a head-mounted speaker assembly and a command and dispatch platform. The Bluetooth microphone is installed in the air respiration mask of the positive pressure air respirator, and is specifically used to collect the voice signals of firefighters. The collected signals are transmitted to the communication host via Bluetooth. The communication host processes the noise through the built-in noise reduction chip. After optimizing the voice signal, it is transmitted to the command and dispatch platform using the 4G network to realize real-time communication between firefighters and the command center. The noise reduction chip loads different artificial intelligence noise reduction algorithms according to different noise scenarios. In response to the background noise at the fire scene, the noise reduction chip loads an explosion noise model based on deep learning. The model removes explosive noise through time-frequency analysis and dynamic filtering algorithms to optimize the clarity of voice signals. By training different noises with artificial intelligence models, the system can adapt to complex noise environments and significantly improve the quality of voice communication. The system also supports voice command control. Firefighters can control the equipment through voice commands without operating the equipment, such as calling the command center or team members at a single point, which solves the need for effective communication when both hands are occupied. Through intelligent noise processing technology and voice command control technology, the communication quality in high-noise environments has been greatly improved, ensuring that firefighters can quickly and accurately convey key information during the rescue process and improving rescue efficiency.

[0039] It should be further explained that firefighters face very noisy environments when performing high-risk rescue missions, such as high-intensity fire noise, equipment operation noise, and breathing noise when wearing positive pressure air respirators. These noises will interfere with the transmission of voice signals and make information communication difficult. Traditional noise reduction technology cannot effectively eliminate all noise and optimize the clarity of voice signals in such an environment.

[0040] Based on noise reduction processing, the intelligent communication system uses adaptive gain control technology to enhance the clarity of speech signals and ensure the intelligibility of speech signals in complex noise environments.

[0041] In this system, the noise reduction chip calculates the ratio of the voice signal to the noise signal by monitoring the amplitude of the audio signal and the intensity of the background noise in real time. Based on the ratio, the system automatically adjusts the gain to improve the clarity of the voice signal. If the background noise is strong, the gain will increase to improve the audibility of the voice signal; when the noise is low, the system will reduce the gain to avoid distortion caused by excessive amplification of the voice signal.

[0042] After adaptive gain adjustment, the system further improves the quality of voice signals through the linear predictive coding (LPC) technology in the voice enhancement algorithm, effectively removing noise and unnecessary components in the voice, and improving the naturalness and comprehensibility of the voice. This system can automatically adjust the gain of voice signals in complex noise environments, such as fire rescue sites, to ensure that firefighters can communicate clearly and stably. At the same time, voice enhancement technology is used to improve voice quality, ensuring accurate and timely information transmission between the command center and firefighters.

[0043] Through the noise reduction chip and the artificial intelligence noise reduction algorithm based on deep learning, the present invention can effectively eliminate the impact of noise generated at the fire scene and when wearing a positive pressure air respirator on the communication quality. Through the application of different noise models and adaptive algorithms, it can remove noise in real time and accurately, and optimize the clarity of the voice signal; through the adaptive gain control technology, the gain is automatically adjusted according to the change of signal strength, thereby enhancing the clarity of the voice signal, ensuring the intelligibility of the voice signal in a complex noise environment, optimizing the quality of the voice, and ensuring that the voice signal can remain natural and clear in a high noise background.

[0044] Through the voice command control function, firefighters can control the equipment without releasing their hands when performing rescue tasks, such as calling the command center or other combat personnel at a single point. This technology significantly improves work efficiency and ensures that firefighters can respond quickly in critical moments.

[0045] Compared with traditional dual-microphone noise reduction technology and bone conduction technology, this invention adopts a noise recognition and noise reduction model based on deep learning, which can dynamically adapt to complex and changeable high-noise scenes. This system not only solves the breathing noise that is difficult to remove with traditional technology, but also provides a more accurate noise elimination effect in high-intensity background noise environments.

[0046] The present invention realizes real-time information interaction between different rescue sites through the 4G network, so that the command and dispatch platform can perform group setting and positioning display between multiple sites. This function provides reliable guarantee for cross-regional coordinated command and information sharing, greatly improving the emergency response capability in rescue missions; through automatic group call mode and intelligent device control, the present invention avoids misoperation caused by frequent manual operation, while reducing the need for shouting and repeated descriptions, extending the use time of positive pressure air respirators, and effectively improving the communication efficiency between firefighters. The efficient communication and equipment control functions of the entire system greatly improve the rescue efficiency and shorten the rescue time.

[0047] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0048] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent communication system and device in a high noise scenario, characterized by: The system includes a communication host, a Bluetooth microphone, a head-mounted speaker assembly and a command and dispatch platform; the Bluetooth microphone is connected to the wearer's positive pressure air respirator and is used to collect voice signals and transmit them to the communication host via Bluetooth. After being processed by a noise reduction algorithm, the communication host transmits the signals to the command and dispatch platform, which is used to receive signals and exchange information. The communication host has a built-in Android operating system, a noise reduction chip and a host chip, which are used to receive audio signals, perform noise reduction processing and signal distribution, and has a built-in terminal APP, which supports automatic login and group calls according to preset groups; the Bluetooth microphone is arranged in the air-exhalation mask of the positive pressure air respirator, and voice collection is performed through a microphone installed in the mask of the positive pressure air respirator, and the sound is picked up and transmitted to the communication host via Bluetooth; the head-mounted sound speaker assembly includes a microphone and a speaker, which transmits voice signals through cables, and makes sounds after noise reduction processing under the processing of the communication host. The microphone is used to pick up sounds and transmit audio signals to the noise reduction chip through cables for algorithm noise reduction processing; the communication host supports voice command control, allowing firefighters to control the equipment through voice commands, including single-point calling of the command center and single-point calling of personnel; information is transmitted through the 4G network, supporting cross-regional real-time communication and information exchange between different rescue sites.

2. The intelligent communication system and device in a high noise scene according to claim 1, characterized in that: The noise reduction chip loads different artificial intelligence noise reduction algorithms according to the set noise scenes. The artificial intelligence noise reduction algorithms are trained on different noise samples based on the deep learning model, adapt to the noise interference in different scenes, and automatically optimize the clarity of the voice signal.

3. The intelligent communication system and device in a high noise scene according to claim 2, characterized in that: The process of the noise reduction chip loading different artificial intelligence noise reduction algorithms according to the set noise scene includes: The process of the noise reduction chip loading different artificial intelligence noise reduction algorithms according to the set noise scene includes: According to the characteristics of the detected high-intensity background noise, the explosion noise model based on deep learning is loaded. This model analyzes the instantaneous frequency characteristics of high-frequency explosion sounds, uses time-frequency analysis and dynamic filtering algorithms to remove explosive noise and optimize the clarity of voice signals. According to the detected low-frequency noise characteristics, a low-frequency noise model based on deep learning is loaded. The model identifies low-frequency noise through spectrum analysis and deep convolutional neural network, and uses low-pass filtering and spectrum subtraction algorithms to remove noise. At the same time, adaptive gain control is used to enhance the clarity of speech signals. According to the detected intermittent high-frequency noise characteristics, a high-frequency noise model based on deep learning is loaded. This model identifies high-frequency noise through periodic noise detection and time-frequency analysis methods, and removes high-frequency noise through bandpass filtering and noise suppression algorithms; According to the characteristics of breathing noise generated when wearing a positive pressure air ventilator, a deep learning-based breathing noise model is loaded. This model identifies the periodic characteristics of breathing sounds, uses spectrum separation and adaptive algorithms to remove breathing noise and optimize the clarity of speech signals.

4. The intelligent communication system and device in a high noise scene according to claim 3, characterized in that: The noise reduction chip loads a deep learning-based explosion noise model according to the detected high-intensity background noise characteristics. The model converts the audio signal using a time-frequency analysis method by analyzing the instantaneous frequency characteristics of the high-frequency explosion sound to generate a time-frequency diagram; the deep learning explosion noise model is trained by a convolutional neural network CNN or a recursive neural network RNN ​​to identify the frequency, amplitude and time characteristics of the explosion noise, and adjusts the parameters of the filtering algorithm according to the recognition results to remove the interference of the explosion noise on the voice signal; the noise reduction chip further includes a dynamic filtering algorithm, which performs adaptive filtering based on the characteristics of the explosion noise and optimizes the clarity of the voice signal; the time-frequency analysis method generates a time-frequency diagram through a short-time Fourier transform STFT or a continuous wavelet transform CWT, monitors the intensity change of the explosion noise in real time, and adjusts the noise reduction strategy according to the instantaneous frequency of the explosion noise. The noise reduction chip also includes a voice enhancement module, which enhances the audibility and clarity of the remaining voice signal through an adaptive gain control technology after removing the explosion noise.

5. The intelligent communication system and device in a high noise scene according to claim 4, characterized in that: The noise reduction chip loads a low-frequency noise model based on deep learning according to the detected low-frequency noise characteristics. The model identifies the characteristics of low-frequency noise through a spectrum analysis method, and uses a deep convolutional neural network (CNN) for training to identify the difference between low-frequency noise and speech signals. The spectrum analysis method generates a spectrum diagram of the audio signal through the fast Fourier transform (FFT) technology to identify the frequency range and amplitude of the low-frequency noise signal. The low-frequency noise model processes the low-frequency noise in the audio signal in real time according to the low-frequency noise characteristics in the training data set, removes the low-frequency noise through a low-pass filtering algorithm, and retains the high-frequency part of the speech signal. The low-frequency noise is removed through a spectrum subtraction algorithm, which subtracts the spectrum of the low-frequency noise in the time-frequency domain to eliminate the interference of the low-frequency noise on the speech. The noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to the change in signal strength.

6. The intelligent communication system and device in a high noise scene according to claim 5, characterized in that: The noise reduction chip loads a high-frequency noise model based on deep learning according to the detected intermittent high-frequency noise characteristics. The model identifies high-frequency noise through periodic noise detection and time-frequency analysis methods, and removes high-frequency noise through bandpass filtering and noise suppression algorithms. The periodic noise detection and time-frequency analysis method performs time-frequency conversion on the audio signal through short-time Fourier transform STFT or continuous wavelet transform CWT, generates a time-frequency diagram, and analyzes the periodic characteristics and frequency fluctuations of high-frequency noise; the deep learning high-frequency noise model is trained through a deep convolutional neural network CNN or a recursive neural network RNN, extracts the characteristics of high-frequency noise from the time-frequency diagram and distinguishes it from the speech signal, and recognizes and suppresses intermittent high-frequency noise in real time; the bandpass filter filters a specific frequency range of the audio signal to remove the high-frequency noise part in the audio signal while retaining the effective frequency in the speech signal; the noise suppression algorithm removes the noise signal overlapping with the speech through spectral subtraction or frequency domain filtering method, and optimizes the clarity of the remaining speech signal through speech enhancement technology; the noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to the change of signal strength, enhances the clarity of the speech signal and ensures the intelligibility of the speech signal in a complex noise environment.

7. The intelligent communication system and device in a high noise scene according to claim 6, characterized in that: The noise reduction chip loads a deep learning-based breathing noise model according to the characteristics of breathing noise generated when wearing a positive pressure air respirator. The model removes breathing noise by identifying the periodic characteristics of breathing sounds and adopts spectrum separation and adaptive algorithms to optimize the clarity of voice signals. The deep learning breathing noise model is trained by a convolutional neural network (CNN) or a long short-term memory network (LSTM) to identify and separate the breathing noise part in the audio signal and distinguish it from the frequency components of the voice signal. The spectrum separation method generates a time-frequency diagram of the audio signal by using a time-frequency analysis technique, and identifies and removes the spectrum part of the breathing noise by using spectrum analysis. The adaptive algorithm dynamically adjusts the intensity of noise removal according to changes in the signal by using Kalman filtering and minimum mean square error (MMSE) to optimize the effective components in the voice signal. The noise reduction chip includes a voice enhancement module to improve the quality of the voice signal by using voice enhancement technology. The noise reduction chip also includes an adaptive gain control module to automatically adjust the gain according to changes in signal strength.

8. The intelligent communication system and device in a high noise scene according to claim 7, characterized in that: The noise reduction chip also includes an adaptive gain control module, which automatically adjusts the gain according to the change of signal strength, enhances the clarity of the voice signal, and ensures the intelligibility of the voice signal in a complex noise environment; the adaptive gain control module calculates the ratio of the voice signal to the noise signal by real-time monitoring the amplitude and noise intensity of the audio signal, and adjusts the gain value accordingly; the gain calculation method includes adjusting the gain value according to the dynamic change of the signal strength, if the noise intensity is large, the gain value is increased; if the noise intensity is small, the gain value is reduced; the gain adjustment process adopts a smoothing control algorithm to avoid distortion caused by sudden gain changes, and ensure the continuity and naturalness of the voice signal; the gain control module is combined with a voice enhancement algorithm to optimize the signal clarity.

9. The intelligent communication system and device in a high noise scene according to claim 8, characterized in that: The voice command control system includes a voice recognition module and a control module. The voice recognition module recognizes and processes voice commands, and the control module performs control operations according to the recognition results.

10. An intelligent communication device in a high noise scene, characterized in that: The invention comprises the intelligent communication system as claimed in claim 1, and a positive pressure air respirator and a command and dispatch platform used in conjunction with the intelligent communication system, wherein: Positive pressure air respirator, which provides breathing support for firefighters, and has a Bluetooth microphone installed on its air respiration mask; The command and dispatch platform is deployed on a tablet or PC, connected to the communication host through a 4G network, and performs group setting, positioning display, single call, group call, temporary group, and broadcast according to the information collected in real time.

Images