11 results about "Simulation noise" patented technology

The invention discloses an intelligent denoising and enhancing method and system for ground penetrating radar signals, and relates to the technical field of ground penetrating radar data processing.The method comprises the following steps that forward modeling software is used for generating a synthetic radar profile, analog noise is added to the radar profile, a noisy profile is generated, and a training data set is constructed; a DRSN-ATT-UNET network model is constructed and trained, the network model takes U-Net as a basic framework, a deep residual shrinkage module is integrated in an encoder, and an attention gating module is integrated in jump connection; actual ground penetrating radar profile data to be processed are loaded, and standardized preprocessing is carried out; inputting the preprocessed profile data into the trained network model, and outputting a de-noised and enhanced radar profile; and carrying out anti-standardization post-processing on the model output, and outputting a radar image with a high signal-to-noise ratio. According to the method, the ground penetrating radar noise in complex environments such as mines can be effectively suppressed, the data quality is remarkably improved, and a reliable basis is provided for accurate geological interpretation.

The application discloses a 12-lead electrocardiosignal denoising method and system based on a dual generative adversarial network, which comprises the following steps: obtaining time-adjacent clean electrocardiosignal segments and noisy electrocardiosignal segments; constructing a noise generative adversarial network, wherein the generator takes a random noise vector as input and simulates noise as output, the discriminator takes the noisy electrocardiosignal segment as a positive sample and a synthetic signal composed of the clean electrocardiosignal segment and the simulated noise as a negative sample for adversarial training, and a noise generator snapshot library is generated; performing dynamic noise injection on the clean electrocardiosignal segment to obtain 'noisy-clean' paired data; constructing a denoising network comprising a denoiser; obtaining the 12-lead electrocardiosignal containing noise and inputting it into the denoiser to output the denoised 12-lead electrocardiosignal. The application aims to overcome the defects in the prior art, such as the lack of real 12-lead electrocardiosignal denoising paired data and the poor model generalization ability caused by excessive dependence on the Gaussian noise assumption.

The application provides a kind of based on the shear wave transformation of Meyer window function's sonar image noise reduction method and system.The method comprises: the transformation of reverberation noise model to Gaussian additive noise model is combined with shear wave transformation;In the process of shear wave transformation, considering the continuity, smoothness and compact support performance shown by wavelet function and scale function in time domain and frequency domain, Meyer window function based on the performance shown in time domain and frequency domain is proposed to construct shear wave filter;In order to avoid the influence of gray overflow in sonar image, and realize better reverberation suppression effect, noise variance estimation based on weak texture block is proposed, and the root mean square coefficient obtained by the shear wave transformation of simulated noise is combined to obtain adaptive filtering threshold in each scale and direction.Experiments show that the method proposed in the application can more effectively suppress reverberation noise, and has better edge preservation effect.

The invention relates to the technical field of defect detection, and provides an industrial defect data generation method based on a visual language large model, and the method comprises the steps: extracting a visual feature vector of an industrial product image and a semantic feature vector described by a defect text through the visual language large model, and mapping the visual feature vector and the semantic feature vector to a unified semantic space; a semantic guidance vector is dynamically generated in combination with a cross attention mechanism; directional simulation noise is generated based on the visual form of the defect, a semantic guide vector is coded into a modulation signal through a language branch of a diffusion model, and deep fusion of noise adding visual features and the modulation signal is achieved in a denoising decoder; the semantic matching degree of the defect image and the defect text description is verified through a visual language large model, the positioning accuracy of a defect area is verified by means of a class activation graph, a verification result serves as a reward signal, diffusion model parameters and weight parameters of a cross attention mechanism are alternately optimized through reinforcement learning, and the defect recognition accuracy is improved. And the finally generated industrial defect data set is stable in quality and reliable in attribute.

The application discloses a kind of ground penetrating radar signal intelligent denoising and enhancement method and system, it is related to ground penetrating radar data processing technical field, including the following steps: using forward simulation software to generate synthetic radar profile, and adding simulated noise in radar profile, generate noisy profile, construct training dataset;Build and train DRSN-ATT-UNET network model, the network model is with U-Net as basic architecture, and integrated depth residual shrinkage module in encoder, integrated attention gate module in jump connection;Load actual ground penetrating radar profile data to be processed, and carry out standardization preprocessing;Profile data after preprocessing is input into the network model that has been trained, and output denoising and enhanced radar profile;After model output is carried out anti-standardization post-processing, output high signal-to-noise ratio radar image.The application can effectively suppress ground penetrating radar noise under complex environment such as mine, significantly improve data quality, and provide reliable basis for accurate geological interpretation.

The present application relates to a kind of bearing degradation system level data generation method based on component level test, it is related to bearing signal measurement technical field, solve the technical problem that must carry out system level test in order to obtain bearing system level degradation signal in prior art.The present application first uses dynamics simulation software or the assembly level simulation noise signal of bearing, then based on bearing component level test device, obtains the component level vibration signal of bearing in full life cycle degradation;Then by extracting high sampling rate, narrow bandwidth under the rotating frequency characteristic signal, obtain the filter signal of simulation signal and component level test signal;On this basis, by the correct peak value that can represent the motion characteristics of bearing obtained by the proposed screening method, finally with these peak values as reference datum, obtain bearing degradation system level data, can avoid the problem that traditional method needs to carry out system overall assembly level test, improve model training data generation efficiency and reduce economic cost simultaneously.

The application relates to a noise adjustment method of an audiometer and a noise generator thereof, comprising the following steps: constructing a noise database, the types of the noise including simulated noise, random noise and mixed noise; acquiring frequency range and amplitude data when pure tone is output, and generating a preselected scheme of masking noise adapted to the pure tone from the corresponding type according to an input instruction, the preselected scheme including the duration, the expected frequency range and the expected amplitude of the selected masking noise; superimposing and outputting the selected masking noise and the pure tone, and acquiring the mixed tone segment signal again at the eardrum, and generating a waveform spectrum through time-frequency conversion; based on the waveform spectrum, removing the waveform spectrum calculated according to the frequency range and the amplitude data of the pure tone, and recalculating the actual frequency range and the actual amplitude of the masking noise, and adjusting the frequency range and the amplitude of the masking noise output according to the actual frequency range and the actual amplitude, so that the actual frequency range and the actual amplitude are infinitely close to the expected frequency range and the expected amplitude respectively.

Automatic exposure with simulated histograms may include obtaining a noise-blur exposure duration value in accordance with a minimal simulated noise-blur cost value obtained in accordance with simulated noise data and simulated blur data, obtaining a saturation exposure value in accordance with a minimal simulated saturation cost value obtained in accordance with simulated black saturation data and simulated white saturation data, comparing the noise-blur exposure duration value and the saturation exposure value, and obtaining a target gain value and a target exposure duration value based on the comparison, wherein the comparing may include obtaining the target exposure duration value in accordance with a minimal simulated blur-saturation cost value obtained in accordance with the simulated blur data and the simulated black saturation data or in accordance with a minimal simulated noise-saturation cost value obtained in accordance with the simulated noise data and the simulated white saturation data.

The application provides an electroencephalogram data synthesis method, device and equipment and a storage medium, and belongs to the field of biological signals. The method comprises the following steps: in response to an electroencephalogram data synthesis task, loading corresponding head model information, wherein the head model information is biophysical basic information; synthesizing electroencephalogram scalp signals according to the head model information and the electroencephalogram data synthesis task to obtain scalp signal data, and outputting electroencephalogram signal causality during the synthesis of the electroencephalogram scalp signals to obtain causality feature data; generating simulated noise for electroencephalogram extraction according to the head model information and the electroencephalogram data synthesis task to obtain noise signal data; and synthesizing electroencephalogram data according to the electroencephalogram data synthesis task, the scalp signal data, the noise signal data and the causality feature data. By adding the electroencephalogram signal causality and the simulated noise for electroencephalogram extraction into the electroencephalogram data, the generated electroencephalogram data has a causal true value and is more accurate.

This invention discloses a method and apparatus for grading the sugar content of kiwifruit by combining multi-scale feature extraction and global inference. Near-infrared spectral and sugar content data from two batches of kiwifruit samples spanning two years are collected as independent datasets. Abnormal spectral data and corresponding sugar content data entries are removed. The remaining spectral data is preprocessed and mixed with simulated noise. Sugar content is graded and used as label values ​​for training the model. A multi-scale attention residual convolutional network is used to extract fine spectral features related to sugar content. A parallel long-distance correlation encoding and decoding model is used to capture the long-pathway dependence of sugar content traits implicit in the spectral data. The trained and tested stable model is deployed to the kiwifruit grading device for online non-destructive testing and automatic grading. This overcomes the problems of low grading accuracy and easy confusion of critical sugar content grades caused by the limited local receptive field in existing technologies, which makes it difficult to capture fine spectral morphology and lacks modeling of long-pathway global dependencies.

This application relates to the field of defect detection technology and provides a method for generating industrial defect data based on a large visual language model. The method extracts visual feature vectors from industrial product images and semantic feature vectors from defect text descriptions using a large visual language model, maps them to a unified semantic space, and then dynamically generates semantic guidance vectors using a cross-attention mechanism. Directional simulated noise is generated based on the visual morphology of the defect, and the semantic guidance vectors are encoded into modulation signals through the language branch of a diffusion model. Deep fusion of the noisy visual features and the modulation signals is achieved in a denoising decoder. The semantic matching degree between the defect image and the defect text description is verified using a large visual language model, and the accuracy of defect region localization is verified using a class activation map. The verification results are used as a reward signal, and reinforcement learning is used to alternately optimize the parameters of the diffusion model and the weight parameters of the cross-attention mechanism, ensuring that the final generated industrial defect dataset is of stable quality and reliable attributes.