82 results about "Stationary noise" patented technology

The invention discloses a self-adaptive noise reduction method for a positive pressure type air respirator based on airflow spectrum characteristics, and relates to the technical field of positive pressure type air respirators.The self-adaptive noise reduction method comprises the steps that a respirator air path pressure signal and an environment noise signal are synchronously collected through an airflow sensor and a microphone array; performing time-frequency analysis on the collected signals, and extracting airflow spectrum features; and establishing a mapping relation between the working state of the breathing machine and the noise spectrum based on the historical record according to the mapping relation between the working state of the breathing machine and the noise spectrum. Through real-time acquisition and analysis of airflow frequency spectrum characteristics, accurate identification of useful airflow signals and environmental noise signals, dynamic adjustment of noise reduction parameters, effective suppression of high-frequency noise and low-frequency vibration, and improvement of noise reduction effect, compared with a traditional fixed frequency band filtering or passive sound insulation material, the method can better cope with a non-stationary noise environment, and has good application prospects. The noise reduction strategy is adjusted in real time, clear auditory information is provided in a complex noise environment, and therefore the auditory perception ability of firemen is improved.

The invention discloses a low signal-to-noise ratio direct spread signal detection method based on noise cancellation, and belongs to the field of communication spectrum sensing. The self-adaptive noise cancellation method based on the minimum mean square error is adopted, non-stationary noise interference can be tracked and eliminated in real time, filtering parameters are automatically optimized in an unknown channel environment, and a signal detection system can adapt to different background noise conditions. And stable signal detection is realized in a low signal-to-noise ratio environment by utilizing a cyclic spectrum analysis method and extracting the cyclic stability characteristic of the direct spread signal. The existence of the signal is judged through the characteristic spectrum peak on the non-zero cyclic frequency, and the carrier frequency and the pseudo code rate are further estimated, so that more accurate signal identification and parameter extraction are realized, the influence of noise uncertainty on the detection performance is avoided, and the detection robustness and reliability are improved. Welch smoothing processing and short-time Fourier transform are combined, the variance of spectrum estimation is reduced when the cyclic spectrum is calculated, and the detection robustness is improved.

The invention relates to a cooling noise control method and device for a sand-proof cooling fan in a high-temperature environment. The method comprises the following steps: measuring and analyzing fan blade parameters, a dustproof net cover structure and temperature-dust environment data, and establishing a high-temperature dust coupling dynamic model considering dust adhesion and temperature deformation; performing cyclostationary characteristic decomposition on the fan operation acoustic signal to obtain an environment sensitive Gaussian mixture cyclostationary noise model; calculating a cyclic spectrum correlation function for the fan noise signal and extracting a feature vector to obtain a noise source identification result under a high-temperature sand and dust working condition; and constructing a heat dissipation efficiency function and a noise level function in a temperature-dust-rotating speed three-dimensional parameter space, executing weight adaptive adjustment calculation, and outputting an optimal rotating speed control strategy. Dynamic balance between heat dissipation requirements and noise control is achieved, and the noise level is effectively reduced on the premise that the heat dissipation efficiency is guaranteed.

The invention belongs to the technical field of voice interaction, and relates to a 5G communication equipment intelligent voice interaction method and system based on deep learning, and the method comprises the steps: generating a joint input signal containing voice frequency domain information and a motion state; outputting the de-noised voice segments and the corresponding environmental interference level parameters; generating a basic voiceprint ID containing a user biological feature identifier and a dynamic key fragment; calculating a cloud processing priority score according to the environmental interference level parameter and the equipment residual electric quantity value, when the score exceeds a preset threshold value, sending a voice processing request containing a basic voiceprint ID to an edge computing node, and otherwise, triggering a local semantic recognition process; and receiving a semantic recognition result from a cloud end or a local end, and generating a multi-round dialogue response instruction in combination with the context parameters in the historical interaction data of the user. According to the invention, the problem that local processing resource overload or cloud communication delay exceeding is easily caused by a fixed noise weight distribution mechanism is solved.

The invention discloses an auxiliary decision-making system for cognitive competence assessment of old people, which relates to the technical field of cognitive auxiliary decision-making and comprises an environmental noise acquisition and feature extraction module, a time sequence fluctuation analysis module, a dynamic threshold reconstruction and judgment module, a cross-domain consistency calibration module, a multi-modal synchronous verification module and a dynamic threshold regulation and control module. According to the method, the environmental noise is dynamically analyzed, so that the defects of a fixed noise suppression threshold and a static feature extraction model in the traditional technology are overcome. A sound field energy distribution model is constructed in real time and time sequence fluctuation analysis is combined so that a noise fluctuation interval can be identified, a voice starting point detection threshold value is dynamically adjusted and evaluation precision is enhanced. Through multi-mode synchronous verification combining a reaction time curve and a facial movement track, a time offset error is corrected, a judgment standard is adjusted in real time according to noise changes, misjudgment and virtual high risk early warning are avoided, and therefore the accuracy and reliability of cognitive ability assessment of the old are improved.

The invention provides a denoising method for non-stationary nonlinear magnetotelluric sounding signals, and relates to the technical field of geophysical exploration, and the method comprises the steps: obtaining magnetotelluric sounding signals from a magnetotelluric depth finder, selecting one from the obtained magnetotelluric sounding signals, carrying out the denoising of the selected magnetotelluric sounding signals, repeating the above operation, and carrying out the denoising of the selected magnetotelluric sounding signals. And carrying out de-noising processing on the other acquired magnetotelluric sounding signals. VMD (variational mode decomposition) and ICA (independent component analysis) are combined, denoising of the magnetotelluric sounding signals is realized from multiple angles of time-frequency domain and statistical property, and when effective signals are interfered by long-time high-energy non-stationary noise, the method can ensure high quality of reconstructed signals, effectively reduce the influence of manual parameter selection, and improve the denoising performance of the magnetotelluric sounding signals. Compared with the prior art, the method has the advantages that the method is high in robustness to noise, clear physical interpretation is provided, limitation of a traditional denoising method is overcome, and a more reliable and accurate data processing means is provided for the field of geophysical exploration.

The invention belongs to the field of time synchronization and data processing, and discloses a satellite and Rowland timing data fusion method and related device.Firstly, original data are intercepted and subjected to mean value removal to eliminate baseline offset, and a frequency domain matrix is reconstructed by combining frequency domain conversion with singular value decomposition; according to the method, periodic term interference signals in a frequency domain are accurately recognized and filtered out through a dynamic threshold strategy, then effective components are reserved through time domain conversion, finally, a noise covariance matrix and observation model parameters are dynamically adjusted based on an adaptive Kalman filtering algorithm, and dynamic weight fusion of double-source data is achieved. By the adoption of the method, the defects that a traditional weighted average method is insensitive in fixed weight, Kalman filtering parameters are rigid and global interference suppression of wavelet transformation is insufficient are effectively overcome, the suppression capacity for non-stationary noise and periodic interference is remarkably improved, fused data have the high precision of a satellite system and the anti-interference characteristic of a Rowland system, and the method is suitable for being applied to the field of satellite communication. And finally, high-reliability and high-stability time synchronization performance is realized.

The invention discloses a gravitational field inversion method considering unsteady noise of a satellite gravity observation value. The gravitational field inversion method comprises the following steps: A, acquiring original observation data of a gravity satellite; b, setting observation value noise as white noise, and constructing a random model considering non-stationary noise; c, performing parameter estimation to obtain a post-test residual error of an observation value; d, estimating the noise auto-covariance and precision factor matrix of the observation value based on the post-test residual error of the observation value; e, repeating the steps C and D for iteration to realize refinement of the gravitational field parameter and observation value random model; and F, evaluating the time-varying gravity field model in the spectral domain. According to the invention, the defects in the prior art can be improved, and the precision and stability of gravitational field inversion are 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 speech enhancement method based on bispectrum nonlinear feature coupling, and belongs to the technical field of natural languages. The method comprises the following steps: performing feature extraction on time-frequency domain representation of input noisy voice through an encoder to generate encoding features; inputting the coding features into a bispectrum feature extractor, extracting high-order nonlinear coupling features through complex frequency spectrum conversion and bispectrum analysis, and fusing the coding features and the bispectrum features through jump connection to form complementary feature expression; an amplitude mask decoder and a phase decoder are adopted to decode the fusion features, and an enhanced amplitude spectrum and an enhanced phase spectrum are predicted respectively; and combining a multi-task loss function with a generative adversarial training mechanism to jointly optimize the performance of the speech enhancement model. According to the method, the speech enhancement quality in a noise environment is remarkably improved, and particularly, higher robustness is shown in a non-stationary noise scene.

A method for noise suppression in a hearing instrument includes using an acousto-electric input transducer of the hearing instrument to generate an input signal from ambient sound. A frequency-band-wise noise suppression is applied to a processing signal derived from the input signal. Stationary noise is detected in the respective frequency band, and depending on the detected stationary noise, an amplification factor of the processing signal is set for the relevant frequency band. An analysis adapted to detect stationary and non-stationary noise is applied to the processing signal. The analysis has a lower frequency resolution than the frequency-band-wise noise suppression and, upon detected presence of stationary and / or non-stationary noise in the analysis of the processing signal, the stationary noise in the frequency-band-wise noise suppression is presumed to be detected in each frequency band, and the corresponding amplification factors are set.

This application belongs to the field of coherent diffraction imaging technology, specifically disclosing an ultra-low dose coherent diffraction imaging method and apparatus. Based on a blind source separation strategy, this application uses principal component analysis to process the acquired diffraction signals, performing noise separation and updating in reciprocal space. This effectively separates mixed noise energy into noise components, thus avoiding crosstalk to the reconstruction process and significantly improving the convergence stability and robustness of coherent diffraction imaging when reconstructing diffraction signals with extremely low signal-to-noise ratios under ultra-low exposure doses. Simultaneously, this application constructs noise components separately for each scanning position and correlates noise components at different positions through low-dimensional spatial projection, achieving non-stationary noise separation. This enables more effective handling of random noise caused by the low quantum efficiency of ultra-short band detectors, thus maintaining extremely high noise robustness and reconstruction accuracy even under ultra-low exposure doses, achieving an effective improvement in resolution.

The invention relates to the field of trachinotus ovatus swarm sound production identification, and discloses a method for identifying sound production of a trachinotus ovatus swarm in a net cage based on dual-channel time-varying Wiener filtering. According to the method, modeling and target signal counteracting are respectively carried out according to the conditions of the left sound channel and the right sound channel of the input signal so as to analyze the noise signal-to-noise ratio characteristics, enhancement of target biological sound and suppression of noise signals are realized in combination with a Wiener filtering method, trachinotus ovatus sound production can be identified in a complex noise environment, and the method has the target positioning capability. According to the method, by introducing a dual-channel signal equalization and interference compensation mechanism and fusing posterior and prior signal-to-noise ratio estimation, the sound production of the trachinotus ovatus swarm can be effectively identified under the complex noise condition without training data, the problems that the performance is poor and a large amount of training data is needed when non-stationary noise is separated by a traditional noise separation method are solved, and the method is suitable for popularization and application. Technical prerequisites are provided for net cage acoustic monitoring work, and possibility is created for rapid, efficient and accurate sound production analysis of subsequent net cage fish schools.

The invention describes a method for noise reduction in a hearing instrument (1), wherein an acousto-electrical input transducer (E1) of the hearing instrument (1) generates an input signal (E1) from ambient sound (2), wherein a frequency-bandwise noise reduction (21) is applied to a processing signal (V1) derived from the input signal (E1), in which stationary noise in the respective frequency band (Bj, Bk) is detected, and depending on the detected stationary noise, a gain factor (Gk) of the processing signal (V1) for the respective frequency band (Bj, Bk) is set, wherein an analysis (12) is applied to the processing signal (V1), which is configured to detect both stationary and non-stationary noise, wherein said analysis (12) has a lower frequency resolution than the frequency-bandwise noise reduction (21).and wherein, in the case of a detected presence (Y) of stationary and / or non-stationary noise in said analysis (12) of the processing signal (V1), the stationary noise is assumed to be detected in the frequency-bandwise noise suppression (21) in each frequency band (Bj, Bk), and the corresponding gain factors (Gk) are set.

The application discloses a kind of deep development tomographic SAR imaging method based on Taylor linearization and classification prior feedback, obtains multiple preprocessed two-dimensional single view complex image data;Constructing city area-oriented tomographic SAR imaging model;ADMM solving framework and augmented Lagrangian function are constructed, and original variable, auxiliary variable and multiplier variable are iteratively updated;After iterative optimization, high-precision city three-dimensional point cloud imaging result is output, and three-dimensional reconstruction of city scene is realized.The application breaks through the limitation that single criterion is difficult to remove high-amplitude false target, effectively suppresses false target caused by non-stationary noise using spatial continuity prior, significantly improves the accuracy of obtaining target height information;By jointly optimizing the Taylor approximation of physical model, the prior guidance of semantic classification and the continuity constraint of spatial geometry, not only the super-resolution ability is improved, but also the smoothness and accuracy of the ground object elevation are ensured while preserving the details of complex structure.

The invention discloses a full-band voice noise reduction method, device and equipment based on a deep neural network. The method comprises the following steps: generating a noisy voice data set according to a preset noise audio file and a preset pure voice file; constructing a full-band noise reduction model, and training the full-band noise reduction model by using the noisy voice data set to obtain an optimal full-band noise reduction model; and using the optimal full-band noise reduction model to perform noise reduction processing on a collected to-be-processed voice signal. According to the technical scheme provided by the invention, the full-band noise reduction model based on the deep neural network is constructed and trained for noise reduction processing, the noise reduction processing of the full-band voice signal is realized, most of common noise in life can be inhibited, the inhibition effect on non-stationary noise is obvious, the overall residual noise is lower, and the noise reduction effect is better. And the quality, definition and the like of the voice after noise reduction are remarkably improved.

The invention relates to the technical field of noise reduction performance evaluation, in particular to a noise reduction performance evaluation method and system for a dynamic signal, and the evaluation method comprises the following steps: extracting a power spectrum from an input dynamic signal, and projecting the power spectrum to a time axis to generate a time domain power sequence; calculating an autocorrelation sequence by the time domain power sequence, smoothing the autocorrelation sequence, and determining the main period intensity of the input signal based on local peak detection; taking the ratio of the main period intensity to the average value of the autocorrelation sequence as PSI; calculating the gradient of the power spectrum in the time and frequency directions, and further obtaining the total gradient amplitude; generating a frequency band mask matrix used for distinguishing the characteristic frequency band and the background area through the dynamic threshold value; according to the frequency band mask matrix and the power spectrum, characteristic frequency band average power and non-characteristic frequency band average power are calculated respectively, and the ratio of the characteristic frequency band average power and the non-characteristic frequency band average power serves as SCS; noise reduction performance evaluation is realized according to the obtained SCS and PSI, the method adapts to a non-stationary noise environment, and the evaluation effect is more comprehensive.

The invention discloses a mute detection method and device, audio equipment and a computer readable storage medium, and the method is applied to the audio equipment, and comprises the steps: carrying out the framing processing of an audio input signal, and calculating the energy characteristic quantity of each analysis frame in a preset time window; intercepting a plurality of time periods in a preset time window, and determining a background noise estimation value corresponding to each time period based on the energy characteristic quantity of the plurality of analysis frames in each time period; obtaining statistical characteristic parameters according to the plurality of background noise estimated values; dynamically generating a mute detection threshold according to the background noise estimation value and the statistical characteristic parameters; and in a preset time condition, comparing the energy characteristic quantity corresponding to the real-time analysis frame with a mute detection threshold value, and judging whether the audio channel to be detected is in a mute state or not. Therefore, the silence detection threshold adaptively reflects the stability or fluctuation of the noise level, and the accuracy and anti-jitter capability of silence detection in the non-stationary noise environment are effectively improved in combination with the preset time condition.

This invention provides a method, device, electronic device, and storage medium for early warning of warp knitting machine faults, relating to the fields of intelligent textile manufacturing and predictive maintenance technology. Through a spindle phase triggering mechanism, heterogeneous data in multimodal observation data are forcibly aligned to the same physical coordinate system, achieving deep semantic alignment of heterogeneous data. A discrete sequence of fault primitives is obtained through a pre-trained feature encoding quantization model, filtering out inherent non-stationary noise during the efficient operation of the target warp knitting machine. Masked fault modeling is performed on the Transformer backbone network during joint training, enabling the pre-trained Transformer backbone network to grasp long-range structural priors in the production process. This allows the Transformer fault prediction model to achieve millisecond-level early warning based on only minor semantic deviations, without waiting for obvious damage to the fabric surface.

The application belongs to the technical field of voice enhancement. The application provides a dual-path lightweight time-frequency domain adaptive neural network model and a use method thereof. The deep separable convolution layer is adopted in the embodiment of the disclosure to reduce the calculation amount of the model and reduce the noise reduction delay of the model. The lightweight Transformer layer proposed adopts an MLP network, replaces the multi-head attention with single-head attention, and greatly reduces the model parameter amount. Time-frequency domain interaction fusion is performed in the middle layer, so that the model can learn more time-frequency domain joint hidden features in the training process, and the noise reduction effect is greatly improved. By introducing a meta-learning dynamic activation function, the dynamic activation function adapts to non-stationary noise, the meta-training framework supports few-shot fine-tuning, and the deployment cost is reduced. In the process of time-frequency feature adaptive fusion, a regularization term is added, so that the model will not excessively deviate to one side of the frequency domain or the time domain during the training process, and the training effect of the model is ensured.

The application discloses a kind of self-adapting noise reduction filtering method, system, medium and equipment, the method includes to input dynamic signal and carry out direct current offset preprocessing;Time-frequency representation matrix is generated by time-frequency transform, and amplitude spectrum is calculated accordingly;Along time axis analysis amplitude spectrum, extract short time and long time noise amplitude and fusion generate basic noise amplitude;Signal noise amplitude ratio is calculated based on basic noise amplitude, and final noise amplitude is determined;According to final noise amplitude, calculate noise reduction spectrum amplitude, retain original phase information, reconstruct as noise reduction time domain signal by inverse transform;The system includes preprocessing module, amplitude spectrum generation module, double time scale noise estimation module, denoising module, noise reduction and signal reconstruction module;The application is accurately described by double time scale noise estimation Dynamic distribution of noise, combined with the adaptive noise reduction adjustment based on signal noise amplitude ratio, effectively suppress complex, non-stationary noise, while enhancing the relative intensity of characteristic frequency band.

The invention discloses a reciprocating water injection pump fault diagnosis method based on multi-scale dynamic noise reduction and a hybrid model, electronic equipment and a storage medium. The method comprises the following steps: acquiring an original vibration signal, a flow value, a pressure value and a temperature value of a water injection pump; performing noise reduction on the original vibration signal to obtain a noise-reduced vibration signal; the characteristic value of the vibration signal is fused with the flow value, the pressure value and the temperature value; inputting an IWOA-BiLSTM model for classification, and outputting fault types of end face wear of the liquid inlet valve and end face wear of the liquid discharge valve; according to an expert mechanism judgment rule, judging a fracture fault of a liquid inlet valve spring and a fracture fault of a liquid outlet valve spring; the expert mechanism judgment rule is combined with the characteristic values of the pressure, the flow and the vibration signal to set or adjust. The method can solve the problems of poor suppression of non-stationary noise and mode aliasing in a traditional method. The problem that a diagnosis model is prone to falling into local optimum is solved, and according to the information one-sidedness of a single parameter, the diagnosis accuracy is greatly improved, and the false alarm rate is greatly reduced.

The invention particularly relates to a voice noise reduction method based on deep spanning feature extraction and feature cross fusion, which comprises the following steps: a double-flow network structure comprising an amplitude feature extraction branch and a phase feature extraction branch is constructed, the amplitude feature extraction branch adopts a multi-scale coding-decoding structure and a time sequence convolutional network for joint modeling, and the phase feature extraction branch adopts a multi-scale coding-decoding structure; local amplitude details are extracted to be dependent on long-range time; the phase feature extraction branch uses a hierarchical decoding structure for recovering multi-level phase structure features. Moreover, a cross-flow feature fusion mechanism is designed between double flows, so that the amplitude features and the phase features are dynamically complementary in the deep network. According to the method, acoustic physical priori and data driving characteristics are combined, noise reduction can be accurately carried out under complex background noise, and the performance of a voice noise reduction task in a non-stationary noise environment is effectively improved; therefore, problems of insufficient time-frequency correlation modeling, insufficient amplitude and phase feature utilization and the like of the existing speech enhancement technology in a complex noise scene can be solved.

The invention discloses a composite noise reduction method based on a seawater desulfurization aeration tank fan. The composite noise reduction method comprises the following steps: S1, deploying a composite noise reduction system; s2, initializing a system; s3, secondary channel modeling: controlling a secondary sound source to play white noise, and collecting response through an error sensor so as to establish or update a secondary channel estimation model; s4, noise reduction operation: starting a fan and operating a composite FXLMS algorithm; and S5, monitoring and optimizing: monitoring the energy value of the error signal in real time, and dynamically adjusting the self-adaptive step size sum according to the convergence state. By combining the feedforward and feedback composite FXLMS structure, the system can process deterministic noise related to reference signals and random noise which cannot be directly measured at the same time, realizes more comprehensive suppression of broadband and non-stationary noise, and is especially suitable for complex sound source characteristics with periodicity and randomness in fan noise.

The application provides a high-dynamic-environment intercom voice enhancement method and system based on intelligent noise reduction. In response to a release event of a push-to-talk button, a two-state noise dictionary is constructed based on impulsive noise components and non-stationary noise components in background noise in a current high-dynamic environment. In response to a press event of the push-to-talk button, when it is detected that there is a transient region matching the impulsive noise components in the noisy intercom audio signal, online updating of the two-state noise dictionary is triggered to obtain an online updating dictionary. The noisy intercom audio signal is reconstructed based on the online updating dictionary to obtain an initial enhanced voice signal. Envelope reconstruction is performed on the voice segment with abnormal zero-crossing rate in the initial enhanced voice signal to generate a final enhanced voice signal. The technical scheme provided by the application can enhance conversation voice in a high-dynamic environment with non-stationary noise and impulsive noise.

The application discloses a single-frequency pulse signal detection method in a complex underwater acoustic environment and belongs to the field of target detection in a complex underwater acoustic environment. The application solves the problem that the existing detection method ignores the unique structural characteristics of target signals and background noise, resulting in low detection accuracy of target pulse signals in a complex marine environment noise background. The application adopts a data-driven method, does not require artificial experience parameters, and realizes linear skeleton feature extraction and structural enhancement through a linear skeleton structure feature extraction module and a graph attention skeleton feature connection enhancement module, effectively extracts the unique feature structure of target signals and background noise, solves the signal detection problem under the conditions of low input signal-to-interference noise ratio and coexistence of stationary noise components and non-stationary interference components in marine environment noise in a complex marine environment noise background, can stably extract and enhance the linear skeleton feature of the target signal in a complex underwater acoustic environment, and significantly improves the accuracy and robustness of target pulse signal detection. The application is mainly applied in a complex underwater acoustic environment.

The present invention provides a denoising method for non-stationary and non-linear magnetotelluric sounding signals, which relates to the technical field of geophysical exploration. The method includes obtaining magnetotelluric sounding signals from a magnetotelluric sounding instrument, selecting one channel from the obtained magnetotelluric sounding signals, denoising it, and repeating the above operations to perform denoising processing on the remaining obtained magnetotelluric sounding signals. By combining variational mode decomposition (VMD) and independent component analysis (ICA), the present invention realizes the denoising of magnetotelluric sounding signals from multiple perspectives of time-frequency domain and statistical characteristics. When the effective signal is interfered by non-stationary noise with high energy for a long time, this method can ensure the high quality of the reconstructed signal, effectively reduce the influence of artificial parameter selection, be robust to noise, and at the same time provide a clear physical explanation, overcoming the limitations of traditional denoising methods and providing a more reliable and accurate data processing means for the field of geophysical exploration.

The invention discloses a radio station voice squelch method, relates to the technical field of radio station squelch, and solves the technical problem that in the prior art, voice and noise are difficult to accurately distinguish under the condition of a low signal-to-noise ratio, and misjudgment is easily caused. The method comprises the following steps: pre-processing an input digital voice signal, performing voice characteristic parameter calculation on the pre-processed digital voice signal, and processing an obtained calculation result in two paths: in one path, directly mapping the calculation result into a voice quality grade through a binary search method to obtain a real-time voice quality grade; the other path performs smooth filtering on the calculation result and then looks up a table through a binary search method to map the calculation result into a voice quality level, so that an average voice quality level is obtained; finally, the real-time voice quality grade and the average voice quality grade are sent to a squelch judgment module for squelch judgment, and squelch switch output is obtained; the method is based on voice characteristic parameter calculation and analysis, and has good adaptability to stable noise.

The invention relates to a voice noise reduction system and method based on spectral subtraction noise reduction parameter optimization, an optimization device is arranged, noise reduction parameters introduced by spectral subtraction are optimized by using a covariance matrix self-adaptive evolutionary strategy algorithm, parameter solidification is avoided, a noise reduction device is further arranged, the noise reduction parameters obtained by the optimization device are substituted into spectral subtraction, and the noise reduction parameters are optimized by using the covariance matrix self-adaptive evolutionary strategy algorithm. And noise reduction is carried out on the input audio through spectral subtraction so as to output the audio signal after noise reduction, so that the parameter combination of the spectral subtraction can be optimized, the optimal condition suitable for audio noise reduction is found, and the adaptability of a non-stationary noise scene is improved.

This invention relates to a keyword detection method in noisy environments based on multi-task joint learning, belonging to the fields of speech signal processing and natural language processing. It extracts multi-layer acoustic representations from noisy signals using a pre-trained audio encoder model, then performs temporal and inter-layer aggregation through global average pooling layers to generate highly compact and noise-resistant shared features, eliminating local interference from transient noise. Finally, the features are input in parallel into detection and classification branches, and a multi-task joint optimization strategy is used for training to enhance the model's ability to distinguish between speech and background noise. A keyword detection test dataset in a real noisy environment is constructed, and a cascaded decision logic of detection followed by classification is adopted in the inference stage to effectively filter invalid inputs. This invention combines the generalization representation power of the pre-trained model with a multi-task collaborative learning mechanism, significantly improving the model's recognition accuracy, system stability, and engineering practicality in low signal-to-noise ratio and real non-stationary noise environments.