87 results about "Transient noise" patented technology

The invention relates to the technical field of carrier communication, and particularly discloses a power line carrier channel simulation method and system based on a dynamic multi-dimensional interference model.The power line carrier channel simulation method based on the dynamic multi-dimensional interference model.The power line carrier channel simulation method based on the dynamic multi-dimensional interference model comprises the steps that a power line noise sample is collected, and characteristic parameters of periodic interference, transient noise and background noise are extracted; classifying and marking according to noise types and corresponding power grid working conditions, and establishing a noise database containing a mapping relation between noise characteristics and working conditions; and based on the noise database, respectively establishing a periodic interference model, a transient noise model and a background noise model, and forming a multi-dimensional noise interference model. According to the power line carrier channel simulation method based on the dynamic multi-dimensional interference model, through multi-dimensional noise modeling and dynamic environment adaptation, the precision, the real-time performance and the scene coverage capability of power line carrier communication simulation are remarkably improved, and the simulation bottleneck in a complex power grid environment in the prior art is solved.

The invention relates to the technical field of industrial laser precision measurement, in particular to a laser centering device for a coupler and a using method of the laser centering device for the coupler. Light spot displacement signals and environment vibration signals are synchronously collected through a laser transmitter and an optical sensor array, and a frequency domain feature decoupling module is adopted to separate a shafting displacement component and an interference component; the time domain correlation verification module is used for calculating signal time sequence correlation and generating a deviation vector, and finally the adaptive compensation decision module is used for analyzing the deviation vector and outputting an adjustment instruction. According to the invention, effective separation of real misalignment and transient noise in a dynamic environment is realized, the technical problem of measurement misalignment of an existing laser centering device in a complex industrial environment is solved, and the coupling installation precision and the equipment operation stability are improved.

The invention discloses a mechanical defect vibration sound feature extraction method for GIS equipment, and belongs to the technical field of equipment detection. Comprising a data acquisition unit, a data transmission unit, a vibration sound processing unit, a data storage unit, a data processing unit and a display reminding unit, and the data acquisition unit is used for acquiring various data required in the use process of the mechanical defect vibration sound feature extraction system of the GIS equipment. The invention provides an optimal correction logarithmic spectrum estimation and Wiener filtering fusion algorithm, noise signals are estimated, the amplitude part in the noisy signals is processed and updated to obtain enhanced sound, the OM-LSA is used for estimating and filtering transient noise, the Wiener filtering is used for filtering steady-state noise, and the noise is filtered. According to the method, original vibration signals and acoustic signals of each operation state of the GIS are fused, a series of characteristic parameters are extracted from a multi-dimensional domain, and effective characterization of the state of the GIS equipment is realized.

The invention discloses a load switching detection method and system based on a dynamic window and anti-interference verification. The method belongs to the technical field of non-intrusive load monitoring. The technical problems that in the prior art, complete event capturing capacity for multi-stage power changes during equipment starting is lacked, and the multi-stage power changes are easily misjudged as multiple independent switching events are solved. Detecting a power abrupt change event through a sliding window bilateral CUSUM algorithm, dynamically adjusting a window overlapping proportion and a judgment threshold according to a power change rate, and distinguishing climbing start and step start; introducing a delay dual-check mechanism, and filtering instantaneous noise interference in combination with steady-state judgment and trend consistency verification; according to the multi-stage power change of the frequency conversion equipment, event classification judgment logic assisted by transient characteristics is designed, and wrong event segmentation is avoided. Through the collaborative design of the dynamic window strategy and the anti-interference verification, the detection precision and robustness of the load switching event in the household power consumption scene are remarkably improved, and the method is particularly suitable for low-frequency sampling and high-noise environments.

The invention provides a passing noise prediction method and device for an engineering vehicle, and relates to the technical field of engineering vehicles. The method comprises the following steps: carrying out sound source division on an engineering vehicle physical model to obtain a vehicle noise source; carrying out position division on a noise passing rapid acceleration area of the engineering vehicle to obtain working condition point locations; performing vehicle parameter calculation based on the vehicle noise source and the working condition point location to obtain rotating speed data and torque data; on the basis of the rotating speed data and the torque data, according to basic data of the engineering vehicle, characteristic data are collected through an acoustic test, and acoustic characteristics of a sound source are obtained; based on the vehicle noise source, noise transfer calculation is carried out according to the engineering vehicle basic data, and a noise transfer function is obtained; and performing transient noise correlation prediction according to the acoustic characteristics of the sound source and the noise transfer function to obtain a passing noise prediction result of the engineering vehicle. The method is an efficient and accurate noise prediction method for the engineering vehicle.

The invention relates to a high-voltage level shift circuit and a layout structure thereof, and the high-voltage level shift circuit comprises a pulse generation module which is used for receiving an input signal and generating a double-path low-voltage domain narrow pulse signal; the level shift module is used for receiving the low-voltage domain narrow pulse signal and converting the low-voltage domain narrow pulse signal into a corresponding double-path high-voltage domain narrow pulse signal; the voltage transient noise detection module is used for detecting voltage transient noise and generating a transient noise intensity signal; the voltage transient noise filtering module is used for filtering voltage transient noise based on the received transient noise intensity signal to generate a double-path high-voltage domain filtering signal; and the RS trigger is used for converting the received high-voltage domain filtering signal into a high-side power device driving signal. The common-mode noise filtering capability and the VS negative bias resisting capability of the voltage transient noise filtering module can be enhanced, the transmission delay of a chip is effectively reduced, and the width of a narrow pulse signal is free of loss in the transmission process, so that the dynamic power consumption is reduced.

The present application relates to pipeline system damping and noise reduction technical field, provide a kind of active noise cancellation pressure relief valve, comprising: valve body, with the liquid inlet cavity and pressure relief cavity being communicated are formed;Bypass pipe, with liquid inlet cavity is communicated;Piezoelectric actuator and pressure relief component, pressure relief component includes adjusting piece, pressure relief elastic piece and pressure relief valve core, pressure relief valve core and pressure relief elastic piece are set in pressure relief cavity, adjusting piece is movably arranged on valve body, piezoelectric actuator is arranged on adjusting piece, pressure relief elastic piece is pressed between piezoelectric actuator and pressure relief valve core;Piezoelectric actuator is configured to drive pressure relief elastic piece, to eliminate the transient noise caused by pulsating pressure when pressure relief valve core moves.The active noise cancellation pressure relief valve provided by the present application introduces active noise cancellation technology into the active noise cancellation pressure relief valve pulsating noise control by arranging piezoelectric actuator, can be adjusted by pressure relief elastic piece The force exerted on the pressure relief valve core can eliminate the transient noise caused by pulsating pressure when the pressure relief valve core moves.

The invention relates to the technical field of digital image processing and computer vision, and discloses a strip detection method based on image data processing, which comprises the following steps: acquiring an original image matrix of the surface of a carrier to be detected; dynamically adjusting diffusion flux according to a pixel neighborhood gradient modulus value, and processing an original image matrix to generate a reference matrix; applying spatial perturbation to the reference matrix and generating virtual residual distribution; performing differential operation on the original image matrix and the reference matrix to construct a feature matrix; and calculating an energy offset vector between the feature matrix and the virtual residual distribution, identifying a non-target noise cluster according to the linearity difference of the energy offset vector relative to a preset displacement amount, and reducing the pixel energy weight of the non-target noise cluster. According to the method, decoupling of physical entity features and transient noise features is realized, and weak damage features are reserved while non-uniform illumination and water mist noise are suppressed.

Systems and methods described herein are directed towards integrating a shield layer into a current sensor to shield a magnetic field sensing element and associated circuitry in the current sensor from electrical, voltage, or electrical transient noise. In an embodiment, a shield layer may be disposed along at least one surface of a die supporting a magnetic field sensing element. The shield layer may be disposed in various arrangements to shunt noise caused by a parasitic coupling between the magnetic field sensing element and the current carrying conductor away from the magnetic field sensing element.

Systems and methods described herein are directed towards integrating a shield layer into a current sensor to shield a magnetic field sensing element and associated circuitry in the current sensor from electrical, voltage, or electrical transient noise. In an embodiment, a shield layer may be disposed along at least one surface of a die supporting a magnetic field sensing element. The shield layer may be disposed in various arrangements to shunt noise caused by a parasitic coupling between the magnetic field sensing element and the current carrying conductor away from the magnetic field sensing element.

The invention relates to the technical field of power equipment on-line monitoring and intelligent diagnosis, in particular to an extra-high voltage transformer multi-modal data conflict fault misjudgment analysis method based on a DS evidence theory and graph enhancement. According to the invention, under the unified time reference and the electromagnetic-thermal physical topology constraint, the consistency fusion of the three-mode data of electromagnetic, acoustic vibration and distributed optical fiber temperature is realized, and the layered suppression and traceable positioning are carried out on evidence conflicts caused by coupling uncertainty and transient noise in the fusion process. Therefore, reliable fault confidence evaluation is output, and an accurate misjudgment traceability basis is provided. A traceable fault confidence coefficient is output in a unified time reference and a physical topology constraint through an electromagnetic-thermal dynamic coupling map which is synchronously acquired by hardware-level three modes and is updated in real time along with working conditions, and a layered D-S evidence conflict suppression and traceability algorithm.

The invention provides a hearing aid intelligent protection device and method integrated on a headgear, and the method comprises the steps: converting an environment sound wave signal into a frequency domain signal through a signal processor in an interlayer of the headgear, and separating a noise dominant frame and a voice dominant frame in a hearing aid; determining a voice masking index between the environmental noise and the voice in the environmental sound wave signal according to a distribution association relationship between the noise dominant frame and the voice dominant frame; according to the energy distribution characteristics of the noise spectrum in the frequency domain signal, determining the noise interference suffered by the hearing aid in the current wearing scene, and determining the transient noise dynamic characteristics of the environment sound wave signal under the current noise interference; and performing disturbance trend prediction on voice signal noise in the hearing aid according to the transient noise dynamic characteristics and the voice masking index, and further updating an adaptive filtering coefficient of the hearing aid according to a prediction result. By adopting the scheme of the invention, the prediction of the noise disturbance trend in a complex and changeable environment can be realized so as to adaptively adjust the noise reduction filter coefficient of the hearing aid.

The invention relates to the technical field of language data processing, and discloses a language data preprocessing method in a multilingual and complex scene, which is a voice data preprocessing system in the multilingual and complex scene based on an AutoPrep framework, integrates five modules, namely a voice enhancement module, a voice segmentation module, a speaker clustering module, a target voice extraction module and a quality filtering module. According to the scheme, differential suppression of steady-state noise and transient-state noise in multilingual voice signals is realized, and particularly in a small language (such as Kazakh and Talanx) scene, the voice signal-to-noise ratio and the language independence of voice features are effectively improved, so that the voice data can be automatically and structurally processed, and the voice signal-to-noise ratio and the language independence of the voice features can be effectively improved. The problem that in the prior art, the phoneme mapping error rate is high due to the fact that small languages lack an exclusive phoneme system processing module is solved, and the availability and the processing effect of low-resource language data are enhanced.

The invention relates to a space-based space small target complex scene dynamic digital imaging simulation method, and relates to the technical field of space imaging simulation. The method comprises the following steps: designing a target imaging model according to simulation system parameters; pixel-level plane drawing is carried out on the target digital library, and a target contour mask of the target under the specific ground pixel resolution is obtained; mapping a fixed star in the simulation view field into the target digital image; generating a target digital image of a diversity space illumination background; sequentially superposing a plurality of transient noises on the original imaging image, and carrying out fixed noise fusion processing to generate a target digital image containing a plurality of noises; and generating a spatial imaging simulation image sequential sequence through comprehensive motion simulation. By designing the target digital library construction method based on the two-dimensional shape feature point description method and the imaging analysis model of the target digital library construction method, general digital modeling can be carried out on various space targets, and the imaging process simulation of a space camera is realized.

The invention relates to the technical field of electrical variable measurement, in particular to an intelligent leakage current real-time monitoring and diagnosis system and method, and the method comprises the steps: obtaining a real-time current signal of a target regional power grid and a corresponding current signal timestamp, and carrying out the intelligent processing of a real-time electric signal, and obtaining a corresponding current signal sequence; performing wavelet packet transformation and intelligent processing on the current signal sequence to obtain a current signal sub-sequence; for each current signal sub-sequence dynamic window sequence set, intelligent feature analysis is carried out on the dynamic window sequence set, and steady-state features and transient-state features of the current signals are obtained; and establishing a current diagnosis model, and performing diagnosis and early warning on the real-time current based on the current diagnosis model. According to the method, the problem of missing detection or false alarm caused by decomposition of a single frequency band, difficulty in accurately distinguishing high-frequency transient noise and low-frequency steady-state abnormity is solved, the threshold value is adjusted according to the actual risk duration, real-time dynamic prediction is realized, and the health management level of a regional power grid is improved.

This invention discloses a preprocessing method for highly polluted pediatric EEG data, belonging to the field of EEG signal processing technology. The invention proposes a robust motion artifact detection method based on a multi-channel voting mechanism. This method overcomes the sensitivity of single-channel detection to transient noise through multi-channel joint decision-making, automatically identifying time periods requiring restoration. A comprehensive preprocessing workflow integrating PCHIP interpolation restoration, adaptive Bayesian wavelet denoising, bandpass filtering, and independent component analysis is designed. This workflow is validated using real pediatric continuous task EEG data. This invention demonstrates excellent performance in preserving frequency band information and improving the weighted signal-to-noise ratio, achieving an average accuracy of 96.4% (205 test cases) and a maximum of 100% based on permutation entropy feature classification. This invention effectively solves the preprocessing challenge of pediatric EEG data in high-noise environments, significantly improving the accuracy and reliability of subsequent classification tasks.

The present application relates to the technical field of new energy vehicle NVH testing, and discloses a range extender transient noise sound quality evaluation method based on physiological signal correction, which constructs a double-branch neural network by using in-vehicle noise audio signals, subject EEG signals and subjective scores under the transient working condition of the range extender, the audio branch outputs the objective benchmark score, and the EEG branch outputs the subjective bias correction amount; the model is trained by using a joint loss function that fuses the score fitting loss, the objective consistency constraint loss and the bias regularization loss; in the inference stage, the objective sound quality score can be output by only inputting the audio features. The present application decouples the noise objective quality and the evaluator's physiological bias, effectively eliminates the interference of individual differences, emotions and fatigue states on the evaluation results, improves the evaluation stability and repeatability, realizes efficient, accurate and robust objective evaluation of the range extender transient noise without relying on EEG, and is suitable for NVH performance testing and optimization in the vehicle development stage.

The invention relates to the technical field of GIS equipment fault diagnosis, and discloses a GIS equipment fault diagnosis method based on vibration-sound fusion, and the method comprises the following steps: S1, carrying out the feature extraction of a sound-vibration signal, namely, extracting a time domain feature, a frequency domain feature, an energy domain feature and a harmonic domain feature; s2, a CatBoost classification algorithm is optimized, wherein the CatBoost is used for processing discrete features through a symmetric line sampling and target coding strategy; and S3, utilizing the CatBoost optimized by the ant lion algorithm: because the classification precision of the CatBoost algorithm is influenced by two input parameters, namely the learning rate m and the tree depth k, introducing the ant lion optimization algorithm to perform optimization calculation on the input parameters. According to the method, an optimal correction logarithmic spectrum estimation and Wiener filtering fusion algorithm is provided, a noise signal is estimated, an amplitude part in the noisy signal is processed and updated, enhanced sound is obtained, transient noise is filtered by utilizing optimal correction logarithmic spectrum estimation, and steady-state noise is filtered by utilizing Wiener filtering.

The invention discloses a method for suppressing switching transient noise of a double-step-down symmetrical half-bridge power converter, and relates to the technical field of power motor driving. Four PCB layout optimization schemes are provided. According to the first scheme, a double-step-down symmetrical half-bridge power converter, four driving loops and four additional auxiliary sources form a vertical layout. In the second scheme, distributed filter capacitors are arranged on the basis of the first scheme; according to the third scheme, a double-step-down symmetrical half-bridge power converter, four driving loops, four additional auxiliary sources and a distributed filter capacitor form a horizontal staggered layout. According to the fourth scheme, two separation grooves are formed on the basis of the third scheme. Based on a circuit-level switch analysis model, a layout method for reducing mutual coupling of parasitic inductors of a power loop is provided, parasitic parameters of a PCB are optimized, transient oscillation of a switch can be suppressed, the precision of output current is improved, and the application requirement under a high-precision occasion is better met.

The invention relates to the technical field of bio-electricity signal processing, and provides a bio-electricity signal anti-interference method, device and equipment and a storage medium, and the method comprises the steps: collecting a bio-electricity signal in real time, and detecting environment noise; filtering the input signal according to the environmental noise and the bio-electricity signal to obtain a target bio-electricity signal; performing short-time Fourier transform on the target bio-electricity signal, and identifying and suppressing stable noise in a frequency domain; and analyzing instantaneous noise of the target bio-electricity signal by using wavelet transform, and removing transient interference in a time domain. According to the method, the parameters of the filter are dynamically adjusted by using the collected environmental noise, so that the interference with the bio-electricity signal can be more accurately removed, stable noise and transient noise are respectively processed through different algorithms of the frequency domain and the time domain, and the problems that a traditional filtering method is effective in stable noise and transient noise processing effect, signal distortion or false detection is easily caused, and the signal quality is poor are solved. And the transmission quality of weak bio-electricity signals is improved.

The present invention discloses an Internet of Things-based power equipment status monitoring system, which relates to the technical field of wireless communication networks. It includes: a spectrum sensing module that detects and identifies the noise characteristics in electromagnetic signals through a quantum spectrum sensing array to obtain corresponding spectrum data and classification results; a dynamic evaluation module that filters out a list of available channels through dynamic channel survivability, and determines corresponding channels for signal transmission from the list of available channels according to the classification results of the SVW classifier; a frequency hopping optimization module that determines a globally optimal frequency hopping sequence through the list of available channels and the total energy between different channels. The present invention realizes the capture of microsecond-level transient noise through a superconducting resonator and a quantum compressive sensing algorithm, and at the same time, real-time filters out anti-interference channels through dynamic channel survivability and obtains an optimal channel combination, enabling it to complete channel switching within 10 ms.

The application provides a kind of intelligent voice interaction system and method, it is related to electronic information technical field, including: acquisition calibration module, for by multichannel acoustic sensor array real-time acquisition mixed audio signal in environment, and set four reference positioning points located at corner point on sensor array, form quadrilateral structure;Based on the area characteristics of quadrilateral generation geometric correction value, and synchronous extraction noise frequency band feature and speech short-time energy distribution characteristics, and utilize geometric correction value to extract the feature and calibrate;Noise identification module is used for based on noise frequency band feature and speech short-time energy distribution characteristics, identify noise type, including wide-band high-intensity noise, low-frequency persistent noise, multi-source human voice interference and burst transient noise, generate label vector, and dynamically adjust noise reduction strategy according to label vector.The application is collected and calibrated audio signal by multichannel acoustic sensor array, improves the accuracy and fluency of voice interaction.

The present invention relates to the field of carrier communication technologies, and specifically discloses a power line carrier channel simulation method and system based on a dynamic multi-dimensional interference model. The power line carrier channel simulation method based on the dynamic multi-dimensional interference model includes: collecting power line noise samples, extracting characteristic parameters of periodic interference, transient noise, and background noise, classifying and labeling them according to the noise type and the corresponding power grid operating conditions, and establishing a noise database containing the mapping relationship between noise characteristics and operating conditions; based on the noise database, respectively establishing a periodic interference model, a transient noise model, and a background noise model to form a multi-dimensional noise interference model. The power line carrier channel simulation method based on the dynamic multi-dimensional interference model in the embodiments of the present invention significantly improves the accuracy, real-time performance, and scenario coverage ability of power line carrier communication simulation through multi-dimensional noise modeling and dynamic environment adaptation, and solves the simulation bottleneck of the prior art in a complex power grid environment.

The invention relates to the technical field of vehicles, in particular to a noise reduction method and device, a vehicle and a storage medium, and aims at avoiding privacy leakage of passengers while reducing noise in the vehicle. The method comprises the following steps: acquiring an in-vehicle sound signal collected by a first microphone array; acquiring a noise signal outside the vehicle, which is acquired by a second microphone array; under the condition that the sound signal comprises a voice signal, sending a first interference signal to the outside of the vehicle through a first loudspeaker array based on the voice signal; the first interference signal is used for covering the voice signal; under the condition that the noise signal is the steady-state noise, executing an algorithm corresponding to the steady-state noise, and generating a second interference signal for counteracting the noise signal; under the condition that the noise signal is transient noise, executing an algorithm corresponding to the transient noise to generate a second interference signal; and sending the second interference signal to the outside of the vehicle through the first loudspeaker array.

The application discloses a micro bearing production and manufacturing detection data analysis method and system based on big data, relates to the technical field of data processing, and pre-trains a recurrent time sequence network through collection of vibration signals of bearings in a known state, obtains an evolution feature vector of bearings in the same batch through the network combined with a transient noise suppression operator, extracts a static feature vector combined with a machining time sequence and an end face image, corrects a visual missing sample after splicing, constructs a reconstructed kernel feature matrix, obtains a feature transformation and a mapping matrix based on two-way alternate optimization of the matrix and a real label, establishes a bearing state database, collects multi-modal data of a bearing to be detected to generate an initial to-be-detected feature vector, calculates the similarity of the to-be-detected feature vector with database samples after mapping, and finally outputs a detection result. The application deeply fuses multi-modal features, effectively overcomes interference such as workshop visual pollution, thermal expansion and group tolerance drift, and greatly improves detection accuracy and robustness.

The invention relates to a pipeline leakage detection method, in particular to a pipeline leakage detection method based on deep learning. The method comprises the following steps: acquiring a pipeline inspection video; capturing a leakage target; comparing the current frame and the historical frame of the leakage target; capturing the change and spatial position of the leakage target in the time dimension, and carrying out feature fusion; carrying out modeling and motion trail prediction on the motion trail of the leakage target; performing feature enhancement on the leakage target after the motion trail prediction to obtain enhanced features; and carrying out detection head monitoring on the enhanced features, and outputting the leakage probability. According to the invention, a temporarily appearing false target and continuously existing real leakage can be intelligently distinguished, the space-time fusion detection mechanism significantly improves the anti-interference capability of the system, and the problem that the transient noise is sensitive and the continuous leakage is slow in the prior art is solved.

The application discloses a pre-processing method based on high-pollution children's electroencephalogram data and belongs to the technical field of electroencephalogram signal processing. The application proposes a robust motion artifact detection method based on a multi-channel voting mechanism, overcomes the sensitivity of single-channel detection to transient noise through multi-channel joint decision, and automatically identifies the repair period. A set of comprehensive preprocessing procedures are designed, which integrate PCHIP interpolation repair, adaptive Bayesian wavelet denoising, band-pass filtering and independent component analysis. The procedures are verified based on real children's continuous task electroencephalogram data. The application performs excellently in preserving frequency band information and improving weighted signal-to-noise ratio, and the average accuracy rate based on permutation entropy feature classification reaches 96.4% (205 test sets), and the highest is 100%. The application effectively solves the pre-processing problem of children's electroencephalogram data in a high-noise environment, and significantly improves the accuracy and reliability of subsequent classification tasks.

The invention provides an active noise reduction method and device, and relates to the technical field of data processing.The method comprises the steps that real-time driving state data of a vehicle are input into a mode recognition model, the current vehicle condition mode category is determined, and the real-time driving state data at least comprise the engine rotating speed; obtaining a target filter coefficient corresponding to the vehicle condition mode category from a preset filter coefficient lookup table; the target filter coefficient is utilized to update a filter coefficient of an adaptive filter in the active noise reduction system, and the adaptive filter generates an anti-noise signal to offset engine noise based on the updated filter coefficient and a reference signal corresponding to the engine rotating speed. The active noise reduction system can respond quickly when the working conditions of the vehicle are switched quickly, and the transient noise reduction performance and the stability of the overall noise reduction effect under the complex driving working conditions are remarkably improved.

A method includes providing, by a signal source circuit of a sensing circuit, a signal to a sensor via a conductor. When the sensor is exposed to a condition and is receiving the signal, an electrical characteristic of the sensor affects the signal. The signal includes at least one of: a direct current (DC) component and an oscillating component. When the sensing circuit is in a noisy environment, transient noise couples with the signal to produce a noisy signal. The method further includes comparing, by a transient circuit of the sensing circuit, the noisy signal with a representation of the noisy signal. When the noisy signal compares unfavorably with the representation of the noisy signal, supplying, by the transient circuit, a compensation signal to the conductor. A level of the compensation signal corresponds to a level at which the noisy signal compares unfavorably with the representation of the noisy signal.

The present application relates to the technical field of speech noise reduction, and discloses a method, device and computer-readable storage medium for noise reduction processing of speech signals. The method comprises: preprocessing a speech signal to be processed to obtain first feature data; processing the first feature data using a preset noise reduction algorithm to filter the stationary noise in the speech signal to be processed and obtain second feature data; inputting the first feature data and the second feature data into a preset noise reduction network to filter the transient noise in the speech signal to be processed and obtain third feature data; determining the speech signal after noise reduction processing according to the first feature data, the second feature data and the third feature data. Through the above method, the preset noise reduction algorithm is used to filter the stationary noise in the speech signal to be processed, and the preset noise reduction network is used to filter the transient noise in the speech signal to be processed, which can combine the advantages of traditional noise reduction and deep learning noise reduction to achieve good noise reduction effect.