63 results about "Noise field" patented technology

The invention relates to the field of adaptive noise reduction, in particular to an adaptive noise reduction method and system for a multi-mode audio SoC main control chip. The method comprises the following steps: acquiring an original audio input signal according to an SoC main control chip, performing time-frequency domain dual deconstruction and cross-frequency domain noise interference structure analysis, and constructing a full-frequency domain noise interference topology table; carrying out multi-mode interference factor separation on the full-frequency-domain noise interference topology table, carrying out multi-noise environment modeling, and constructing a real-time noise scene model; time sequence noise slope fluctuation modeling is carried out on the real-time noise scene model, dynamic noise reduction response learning optimization is carried out, and a dynamic noise reduction optimization strategy is constructed; and performing neural coding gain and audio boundary line reconstruction according to the original audio input signal to obtain a coding gain key audio signal and a weak audio optimization signal. According to the invention, by flexibly adjusting the noise reduction intensity of the audio signal, the real-time scene noise reduction performance is optimized.

The invention discloses a noise reduction processing method and device, a vehicle and a storage medium, and relates to the technical field of intelligent cabins, and the method comprises the steps: carrying out the recognition based on an error signal and an accelerometer signal of the vehicle, and obtaining a noise scene where the vehicle is located; decomposing the accelerometer signal into a plurality of sub-band signals which are not overlapped with one another; for each sub-band signal, determining a noise reduction processing parameter corresponding to each sub-band signal based on the energy analysis and noise scene of the sub-band signal; based on the noise reduction processing parameter corresponding to each sub-band signal, performing adaptive filtering on each sub-band signal in combination with the error signal to obtain a noise reduction signal corresponding to each sub-band signal; and carrying out fusion reconstruction on the noise reduction signals corresponding to the plurality of sub-band signals to obtain a full-band noise reduction signal. According to the invention, adaptive filtering and noise reduction are carried out on each sub-band signal based on the noise scene where the sub-band signal is located, efficient division and treatment and response can be carried out on complex and multi-spectral structure noise, the noise reduction effect is improved, and the vehicle use experience feeling of a user is further improved.

The invention discloses an audio signal measurement method and device, an electronic device and a storage medium, and relates to the field of audio signal measurement, and the method comprises the steps: deploying a detection microphone nearby a tested device, and arranging a reference microphone in a noise field, so as to synchronously collect a mixed signal containing a test signal and noise, and a pure ambient noise reference signal. After the noise of the two signals is subjected to correlation evaluation and the consistency of noise sources is confirmed, dynamically changing noise components are accurately stripped from the signals of the tested equipment by utilizing real-time noise data acquired by the reference microphone and through an adaptive noise cancellation algorithm, so that pure test signals are restored. Therefore, the problem of non-stationary noise pollution is solved through an active signal processing means instead of depending on imperfect physical sound insulation, and accurate and efficient measurement of the performance of the microphone in a noisy environment is realized.

The invention relates to a DAS signal positioning method based on adaptive tensor decomposition and dynamic correction. The method comprises the steps of converting phase difference data into strain rate data, constructing a three-dimensional tensor based on a time domain signal of a space point, solving an optimization problem after constraint modeling, and extracting a de-noised DAS signal. Calculating an initial fault position, and calculating an actual optical path and an apparent position after temperature change based on a thermal expansion effect and a thermo-optic effect; and a dynamic correction algorithm is set, a system coordinate reference is adaptively calibrated, and a final positioning result is calculated. According to the method, original signals are separated into a low-rank background field, a sparse event field and a structured noise field through unsupervised tensor decomposition, and the defects of dependence on labeled data and insufficient complex noise separation are overcome; by establishing a temperature-optical path coupling physical model and a dynamic correction algorithm for real-time cross-correlation calibration, positioning drift caused by environmental factors is accurately compensated, and high-fidelity denoising and accurate positioning of cable line events in a complex environment are realized.

The invention provides an unmanned aerial vehicle identification and tracking method and system based on multi-modal fusion and trajectory modeling, and relates to the technical field of target detection and tracking. Infrared / visible light images and audio signals are synchronously collected in an airport key area, visual and acoustic features are extracted through a deep network after time alignment, and the target detection and tracking accuracy is improved. Outputting respective unmanned aerial vehicle category probabilities and candidate target frames; afterwards, weights are distributed adaptively according to confidence coefficients of audio and visual classification results, category fusion is carried out on a probability level, and unified multi-modal fusion representation is constructed on a feature level, so that relatively high recognition reliability is still kept in a small-target, weak-texture and strong-noise scene; on the basis, the multi-modal fusion feature sequence serves as input, the position and speed of the unmanned aerial vehicle in each time step are constructed into a trajectory state sequence, a sequence generation model outputs a complete trajectory in an autoregressive mode under conditional constraints, and continuous tracking, occlusion interval complementation and future trend prediction of the motion of the unmanned aerial vehicle are achieved.

The invention relates to the technical field of acoustic metamaterial, in particular to an acoustic metamaterial integrating vibration reduction, sound insulation and sound absorption and a preparation method of the acoustic metamaterial. The acoustic metamaterial capable of giving consideration to vibration reduction, sound insulation and sound absorption comprises a sound absorption functional layer, a sound insulation functional layer and a vibration reduction functional layer, the sound absorption function layer adopts a series-parallel Helmholtz resonant cavity coupling porous material and is used for realizing a broadband sound absorption effect; the sound insulation functional layer adopts a polyurethane filling gradient gradual change minimal curved surface and is used for realizing a broadband sound insulation effect; the vibration reduction function layer is coupled based on a band gap mechanism and used for achieving the broadband vibration reduction effect. The function of comprehensively controlling the environmental noise by means of vibration reduction, sound insulation and sound absorption can be achieved, the efficient control capability on the environmental noise is achieved, the design and manufacturing complexity is low, and the device can be conveniently used in various noise occasions.

The application discloses a rotating machinery fault diagnosis method based on time-frequency domain joint modeling and self-supervised learning, and belongs to the technical field of rotating machinery state monitoring and fault diagnosis, and comprises a vibration signal preprocessing module, a time-frequency domain joint feature modeling module, a self-supervised learning diagnosis module, a confidence check closed loop module and an incremental iteration optimization module; the vibration signal preprocessing module is used for receiving a raw vibration signal of rotating machinery, performing signal standardization preprocessing, generating three types of input data in a time domain, a frequency domain and a time-frequency domain, adaptively weighting fault sensitive dimensions through a channel attention mechanism, comprehensively representing fault evolution information, significantly improving fault feature distinguishability in a variable working condition and a strong noise scene, and guaranteeing the identification ability of early weak faults.

The invention provides a cable joint partial discharge fiber bragg grating sound wave intelligent sensing and detection method, and relates to the technical field of cable partial discharge detection, and the method comprises the steps: obtaining multi-channel time-domain spectral response data of a detected cable joint through a fiber bragg grating sensing array, and demodulating the data to obtain a multi-dimensional feature vector group; decomposing the waveform signal into narrowband intrinsic mode components, extracting energy density through Hilbert transform and executing singular value decomposition, and screening dominant mode components to reconstruct the waveform signal to obtain stress wave time sequence characteristics; calculating a waveform high-order cross-correlation tensor as a hyperedge weight, extracting a time-frequency amplitude feature vector as an initial node feature, constructing a space-time hypergraph, and generating a node embedding vector through graph convolution; and initializing a Gaussian noise field based on a node embedding vector and setting condition information, executing reverse sampling denoising to generate a three-dimensional space probability density field to identify an initial positioning coordinate, and iteratively correcting the positioning coordinate and a wave velocity value until a residual error converges.

The invention discloses an ocean three-dimensional temperature and salt field reconstruction method and system, electronic equipment and a medium. The method comprises the steps that a target field and an observation field are constructed; sequentially adding multiple levels of Gaussian noise to each piece of sample data in the standardized target field, and constructing a noise field sample data set; constructing an ocean three-dimensional temperature and salt field reconstruction model comprising a convolution local dependence modeling module, a patch division module, a multi-head attention global dependence modeling module, a patch restoration module and a convolution reconstruction module; training the ocean three-dimensional temperature and salt field reconstruction model by adopting the noise field sample data set to obtain a trained ocean three-dimensional temperature and salt field reconstruction model; and denoising the pure Gaussian noise field by adopting the ocean three-dimensional temperature and salt field reconstruction model which is guided and trained by the observation field to obtain an ocean three-dimensional temperature and salt field reconstruction result, and carrying out de-standardization on the ocean three-dimensional temperature and salt field reconstruction result to obtain a reconstructed ocean three-dimensional temperature and salt field. According to the invention, the accuracy of ocean three-dimensional temperature and salt field reconstruction can be improved.

The invention relates to the technical field of unmanned aerial vehicle acoustic measurement, in particular to a method for acoustically measuring environmental noise of unmanned aerial vehicle mounting equipment, which comprises the following steps of: during the flight of an unmanned aerial vehicle, continuously acquiring original noise signals, and synchronously recording rotating speed, attitude, acceleration and track position data; generating a harmonic sequence according to the rotating speed of the rotor wing, establishing a rotor wing echo template, and adjusting template parameters in combination with a track position and a reflection path; carrying out difference on the original noise and the rotor echo template in a time-frequency domain to obtain an echo residual signal, and carrying out local interpolation on the gap to form a continuous first noise sequence; identifying a wind shear event segment and reconstructing the event segment according to front and back normal segment trends to obtain a second noise sequence; a sound pressure consistency entropy value is calculated in a multi-track intersection area, and a minimum entropy measuring point is selected as a calibration anchor point; and carrying out trend correction on the adjacent track sequences by taking the anchor points as references, constructing a measurement point network diagram, fusing multi-track data, and finally obtaining a noise field of the target area.

The invention relates to the technical field of artificial intelligence, and discloses a text data preprocessing method suitable for a large financial model, and the method comprises the steps: receiving an original financial text sequence sorted according to timestamps; performing word segmentation and part-of-speech tagging; dynamically calculating an emotional fluctuation index of each text unit based on the sliding time window; self-adaptively setting an emotion noise judgment threshold value and marking candidate noise units; semantic stability is evaluated through embedding space cosine similarity, and noise is confirmed for the second time; after the confirmation noise is removed, anaphora resolution and logic connection reconstruction are executed to repair context coherence; and outputting the pure text sequence. The system comprises corresponding function modules. According to the method, through a dynamic threshold and semantic dual verification mechanism, the effective information retention rate and the noise elimination accuracy rate are remarkably improved in a high-noise scene, and the quality and the processing throughput of financial large model input data are guaranteed.

The application discloses a voice wake-up method, a training method of an acoustic model and related devices. The voice wake-up method comprises the following steps: receiving a voice signal to be recognized, and obtaining an acoustic feature of the voice signal to be recognized; inputting the acoustic feature into a trained acoustic model to obtain a probability of a wake-up word; wherein the step of training the acoustic model comprises the following steps: constructing an enhanced model, the enhanced model comprising a first branch which is the same as the acoustic model; training the enhanced model and the acoustic model together by using training data, and synchronously updating and keeping the same parameters of the first branch and the acoustic model; and determining whether to wake up based on the probability. In the foregoing manner, the application can improve the frame classification accuracy of the acoustic model in a noise scene, thereby greatly improving the wake-up rate in the noise scene.

The invention provides a method for analyzing a high-frequency aerodynamic noise field in a closed space, which comprises the following steps of: 1) based on a CAD (Computer Aided Design) model, carrying out computational domain cutting and grid division on a flow field outside the model, dividing the flow field into a sound source area and a noise radiation area, and carrying out local grid encryption on the sound source area to ensure the subsequent turbulence resolution; 2) solving a flow field by adopting a DDES method, and capturing near-field sound source pulsation; 3, the generalized Green function analysis form in the limited space is introduced into an FW-H equation for integration, and far-field noise considering wall and ground boundary reflection is output. The concept is reasonable, the prediction accuracy, comparable with APE, in the tunnel can be obtained, and a new feasible way is provided for rapid evaluation and optimization of high-speed train tunnel acoustic performance.

A machine learning network is trained to generate noise field images from radiographic (x-ray) images. The training includes accessing a number of previously acquired radiographic images, duplicating the previously acquired radiographic images, and conditioning each of the duplicated images with simulated noise content to form a plurality of simulated low-exposure images. Each of the simulated low-exposure images is paired with its corresponding previously acquired image to form a learning pair. The machine learning network is trained to generate noise field images using the learning pairs of images. A noise-suppressed image of a subject can be generated by applying a scaling factor to at least a portion of a corresponding noise field image and combining the scaled noise field image with a currently captured image of the subject.

This invention relates to a Bayesian estimation-based method for suppressing cross-polarization SAR noise, comprising: acquiring raw cross-polarization data and preprocessing it; acquiring noise vector information corresponding to the cross-polarization data and stripe boundary information of each sub-band; calculating the noise scaling factor of the sub-band based on the Bayesian method; calculating the power balance factor of the sub-band; and subtracting the reconstructed two-dimensional noise field from the backscattering coefficient of the raw cross-polarization data to achieve denoising. The beneficial effects of this invention are: it reads and calculates noise vector information from data products, obtains noise scaling factors through Bayesian estimation, and scales the noise vector to achieve the purpose of suppressing azimuth noise.

This invention discloses a polarity testing method for metering devices suitable for high-noise environments, belonging to the field of metering device testing technology. The method includes: generating a square wave excitation signal with preset parameters based on algorithm control and applying it to the primary side of the metering device; acquiring the secondary side induced response signal, which is then synchronously transmitted after filtering, baseline drift correction, and amplitude scaling preprocessing; integrating the excitation and preprocessed signals, extracting phase difference, delay time, and amplitude characteristic parameters, and sequentially performing polarity, phase sequence, and on / off determination to obtain a preliminary comprehensive result; based on the preliminary result and characteristic parameters, outputting the determination result and confidence level through an improved CNN model, and obtaining the final polarity conclusion through weighted fusion; and visually outputting the final conclusion. This invention, through the collaborative design of precise excitation, anti-interference preprocessing, multi-dimensional preliminary determination, and AI-fused final determination, effectively suppresses noise interference, improves the accuracy and stability of polarity determination, and is suitable for high-noise scenarios such as industrial sites and substations.

The invention relates to the field of float glass production, in particular to a method and a device for identifying glass fragmentation sound in an annealing kiln. According to the method, a plurality of acoustic sensors which are arranged on a longitudinal steel beam on the non-transmission side of the annealing kiln and can tolerate the high temperature of 100 DEG C or above are used for collecting field environment sound signals; sequentially executing time domain preprocessing, improved spectral subtraction noise suppression, wavelet packet analysis and scale energy feature extraction to obtain a normalized feature vector; and finally, carrying out model training and real-time identification based on a hidden Markov model, and outputting whether the sound is glass fragmentation sound or not and a specific fragmentation type. The invention further provides a device for implementing the method, manual guarding can be replaced, accurate recognition and type judgment of glass fragmentation of the annealing kiln are achieved, the labor cost is reduced, manual monitoring defects are avoided, the fragmentation interval can be rapidly positioned, the device is adaptive to the high-temperature and high-noise scene of the annealing kiln, and the yield loss and the equipment production halt risk are reduced.

The invention discloses a voice interaction graphene AI intelligent tea table control method, and relates to the technical field of home voice interaction, and the method comprises the control steps: S1, monitoring a surrounding user instruction based on a built-in microphone of a tea table, and after the tea table receives the voice instruction, synchronously carrying out the preliminary noise reduction processing, and retaining the voice segment of the instruction; s2, training a data set containing a noise scene, enabling a built-in model to learn features for distinguishing voice and noise, and filtering non-voice signals in real time; and S3, for dialects and accents in different regions, collecting a scale data set, training a dialect recognition model, and inputting the obtained voice signal into the dialect recognition model to recognize and judge the current dialect and accents. Through preliminary noise reduction and microphone array layout, the voice acquisition quality is improved, and instruction omission is reduced; secondly, by training a multi-noise scene data set, the adaptability of the model to a complex environment is enhanced, and reliable recognition is guaranteed; and aiming at dialects and accent optimization, the product universality is improved.

The invention discloses a high-precision denoising method suitable for gravity data of a metal mine area, which belongs to the technical field of gravity data denoising, and comprises the following steps: (1) data preparation and initial preprocessing; (2) quickly focusing, inverting and extracting a geologic structure template; (3) forward modeling calculation and signal extraction; (4) noise field estimation and feature analysis; (5) constructing an adaptive filter; and (6) applying filtering and iteration to terminate judgment. According to the method, the inherent limitation of an existing frequency filtering method in the aspect of distinguishing mine-induced anomalies and noise is effectively solved, the dependence on artificial experience is remarkably reduced, the objectivity of filtering parameter selection is improved, a reliable scheme is provided for high-precision denoising of metal mine area gravity data, and the method is suitable for popularization and application. And effective support is provided for accurate detection and positioning of the weak concealed ore body.

The invention discloses an evolutionary multitask-based sonar data noise robust anomaly detection method, which comprises the following steps of: 1, preprocessing acquired sonar data, and constructing a noise-containing data set in a label overturning manner; 2, screening out noise-free samples by using a local outlier factor algorithm; 3, the number of iterations is set, and populations are initialized on the noise-containing data and the noise-free data respectively; 4, taking a true positive rate TPR and a false positive rate FPR as objective functions; 5, performing iterative optimization on the two populations by adopting an evolutionary multi-task algorithm; and 6, taking an optimal noise individual in the noise sonar population as a sonar noise robust classification model to realize classification of target sonar data. According to the method, the abnormal sample can be identified in a noise scene, and a reliable classification result is obtained, so that the robustness of sonar data anomaly detection can be improved.

The application discloses a cross-modal silent speech reconstruction method and system based on ear canal air pressure micro-motion sensing, and belongs to the technical field of human-computer interaction and wearable computing. The method uses a micro-pressure sensing unit placed in an in-ear earphone to collect a non-acoustic air pressure sequence caused by the movement of a sound-producing organ; through adaptive baseline drift suppression and rhythm perception data enhancement processing, a robust feature space is constructed; further, an end-to-end deep neural network containing domain adversarial adaptation, cross-modal semantic alignment, coarse-grained mel-spectrogram generation and residual detail correction is used to map the TPVS to a high-fidelity acoustic mel spectrum. The application effectively breaks through the technical bottleneck of the lack of high-frequency acoustic features in low-frequency mechanical signals, realizes high-precision silent speech command analysis in a mobile and noisy scene, and introduces a coupled quality evaluation gate and trigger-based start / stop control at the inference end to suppress invalid inference and reduce power consumption when wearing is poor or there is no trigger condition.

The invention provides a sound insulation cover field sound insulation effect evaluation method based on oil tank vibration and sound signal inversion, and relates to the technical field of substation reactor sound insulation cover noise field evaluation. The method comprises the following steps: firstly, collecting surface vibration data of the extra-high voltage shunt reactor, and calculating an average sound pressure level of body noise according to the surface vibration data; constructing a multi-sound-source noise calculation model, and combining the frequency domain data of the full-band noise of the receiving point to solve the sound propagation attenuation under each frequency; a sound source single-band sound power inversion model is further established, and the sound power and the full-band sound power spectrum of the sound source are obtained; and finally, the insertion loss of the acoustic enclosure is calculated according to the full-band sound power spectrum and the average sound pressure level of the body noise, and accurate quantitative evaluation of the field sound insulation effect of the acoustic enclosure is realized. The method can effectively overcome the defects that traditional measurement is easily interfered by environmental noise and real insertion loss cannot be obtained.

The invention discloses a far-field speech enhancement video call system, and relates to the technical field of speech enhancement. Comprising a millimeter wave radar module, a microphone array module, a track processing and feature extraction module, a sound source position prediction and positioning parameter conversion module, and a correlation model and dynamic compensation module. A millimeter-wave radar tracks a user movement track in real time, a user position is associated as a sound source position, and initial positioning accuracy is ensured in combination with a Doppler effect. The sound source position at the next moment is calculated in advance by a linear prediction algorithm based on a historical track, and is converted into a look-ahead positioning parameter, so that the positioning delay is reduced. A correlation model is constructed by extracting track displacement and speed characteristics and voice amplitude and frequency proportion characteristics, dynamic compensation is started when the moving speed of a user exceeds a preset threshold value or a risk index exceeds a threshold value, beam forming weight is optimized, and response time is controlled within a certain period of time. The anti-interference capability is improved through multi-modal data fusion, and the method is adaptive to mobile and strong-noise scenes.

The invention discloses a position determination method and device of noise reduction equipment and a nonvolatile storage medium. The method comprises the following steps: setting initial positions and speed ranges for a plurality of active noise reduction devices in a target area; establishing a fitness function; on the basis of the initial position, the speed range and the fitness function, the multiple active noise reduction devices are controlled to move, the adjusted positions corresponding to the multiple active noise reduction devices are determined, and the noise reduction effect after the positions are adjusted is determined; and repeating the operations of controlling the plurality of active noise reduction devices to move, determining the adjusted positions corresponding to the plurality of active noise reduction devices and determining the noise reduction effect after the positions are adjusted until a preset stop condition is met, and determining the target positions corresponding to the plurality of active noise reduction devices in the target area. The technical problem that the noise reduction effect is poor due to the fact that an existing noise reduction sensor is usually placed at a fixed position and cannot be dynamically adjusted according to noise field changes is solved.

The application discloses a kind of air conditioner air inlet sound attenuation structure and the middle double-layer flow air conditioner with the sound attenuation structure, including air conditioner box shell assembly, the air conditioner box shell assembly side has air inlet, the air conditioner box shell assembly is installed with sound shield in the position corresponding air inlet, the upper side of the sound shield is open, so that airflow can be introduced into the air inlet from top to bottom.The beneficial effects of the present application are: after the air conditioner box is installed in the interior of automobile instrument desk, the sound shield can change the air inlet direction of airflow, so that the airflow is introduced into the air inlet from top to bottom, ensure that the air noise field emitted when airflow enters the air inlet is not towards the direction of main and deputy driver, to achieve the purpose of active noise reduction, so as to improve the air conditioner NVH performance.

The invention discloses a cross-modal silent voice reconstruction method and system based on ear canal air pressure micro-motion sensing, and belongs to the technical field of man-machine interaction and wearable computing. The method comprises the following steps: acquiring a non-acoustic air pressure sequence caused by vocal organ movement by using a micro air pressure sensing unit arranged in the in-ear earphone; constructing a robust feature space through adaptive baseline drift suppression and rhythm perception data enhancement processing; and further mapping the TPVS into a high-fidelity acoustic Mel spectrum by using an end-to-end deep neural network including domain adversarial adaptation, cross-modal semantic alignment, coarse-grained Mel spectrum generation and residual detail correction. The technical bottleneck that low-frequency mechanical signals lack high-frequency acoustic features is effectively broken through, and high-precision silent voice instruction analysis in a mobile and noise scene is achieved; and coupling quality evaluation gating and trigger type start / stop control are introduced at a reasoning end, so that invalid reasoning is inhibited and power consumption is reduced under the condition of poor wearing or no trigger.

The invention provides a bearing fault lightweight diagnosis method for an industrial noise environment, and the method comprises the steps: collecting a vibration signal of a bearing, and dividing the vibration signal into a plurality of vibration signal samples through a sliding window; the lightweight bearing fault diagnosis network comprises an input layer, a feature extraction layer and a classification layer which are connected in sequence, and the feature extraction layer comprises at least one enhanced phantom module; training a lightweight bearing fault diagnosis network by using the vibration signal samples; and inputting a to-be-detected vibration signal acquired in real time into the trained diagnosis model for online reasoning, and outputting a fault diagnosis state of the bearing. For an industrial strong noise scene, a multi-scale anti-noise network is further constructed based on a lightweight bearing fault diagnosis network, and model noise robustness is enhanced through multi-scale feature extraction and anti-noise training strategy optimization. According to the invention, double challenges of model lightweight and noise robustness in edge device deployment are solved, and high-precision and high-efficiency fault diagnosis is realized.

The invention relates to a PINN noise field continuation-based seabed reflection loss high-precision estimation method, which comprises the following steps of: 1, giving an N-element vertical receiving array, the depth of each array element and noise field data of a frequency point received by the N-element vertical receiving array; 2, dividing M distance grids according to the frequency and the computational domain for processing the noise field data, and step 3, designing a physical information neural network (PINN) for noise field data continuation, and training the network by taking the collected noise field data as input to obtain a trained NoisePINN network. According to the method, the limitation of vertical receiving array space sampling is effectively broken through, the vertical array aperture is equivalently expanded, and a seabed reflection loss result with higher precision and resolution can be obtained.

The invention discloses an equipment part fault diagnosis method based on a noise residual fusion strategy, and relates to the technical field of fault diagnosis, and the method comprises the steps: obtaining noise-containing state monitoring data collected when mechanical equipment operates in a noise scene, and carrying out the standardized preprocessing of the noise-containing state monitoring data to unify the data distribution and eliminate the dimensional difference; and inputting the preprocessed noise-containing data into the trained denoising-diagnosis combined model, filtering noise through the signal denoising model to obtain denoised data, integrating noise residual errors and the denoised data into comprehensive fusion features through the fusion module, and completing fault prediction through the diagnosis model. Noise interference is effectively suppressed through the signal denoising model, the fusion module avoids loss of important fault information, the joint model realizes denoising and diagnosis collaborative optimization through cascade logic, the problems of poor collaborative effect and easy information loss in the prior art are solved, the robustness and accuracy of fault diagnosis in a noise scene are improved, and the fault diagnosis efficiency is improved. The method is suitable for fault monitoring and diagnosis of various industrial devices.

The invention discloses a self-adaptive voice noise reduction method and system in a multi-noise scene, and relates to the technical field of voice noise reduction. Comprising the following steps: S1, preprocessing input voice data, operation control training support data and wavelet analysis configuration data; s2, carrying out cross-scale structure disassembly and direction texture decomposition analysis, and carrying out prior quality constraint; s3, constructing a structural form consistent with the internal feature map of the deep network; s4, executing a structure modulation and detail coupling mechanism; s5, constructing a loss constraint system and a hierarchical supervision structure; and S6, carrying out coefficient adjustment, strategy switching and network path control. The problems that for conversation conferences and noisy environment scenes, existing noise reduction is prone to generalization and failure under non-stable noise and environment equipment changes, global energy and high-frequency details are not sufficiently considered, self-adaptive control is lacked due to time delay and computing power constraint, and consequently definition and audible identifiability are difficult to stably reach the standard are solved.