51 results about "Time–frequency representation" patented technology

The invention provides a sound wave signal self-supervised learning enhanced robust noise reduction processing method and system, and relates to the technical field of signal processing, and the method comprises the steps: carrying out the feature enhancement of an initial time-frequency representation through a dynamic adaptive mask strategy, constructing a self-supervised reconstruction task based on the mask time-frequency representation, separating noise and signal subspaces in a semantic manifold space, and carrying out the self-supervised learning enhanced robust noise reduction. And establishing a probability propagation network in combination with time domain continuity characteristics to model a local dependency relationship, and finally generating a noise reduction weight and realizing semantic fidelity optimization. The method can effectively improve the noise reduction effect and semantic integrity of sound wave signals in a noise complex environment.

In accordance with implementations of the subject matter described herein, a solution for speech enhancement is proposed. In this solution, a target time-frequency representation at least indicating intensities of an input audio signal at different frequencies over time is obtained. The input audio signal comprises a speech component and a noise component. Frequency correlation information and time correlation information of the input audio signal is determined based on the target time-frequency representation. A target feature representation is generated based on the frequency correlation information, the time correlation information, and the target time-frequency representation. The target feature representation is for distinguishing the speech component and the noise component. An output audio signal is generated based on the target feature representation and the target time-frequency representation. The speech component is enhanced relative to the noise component in the output audio signal. In this way, the performance of speech enhancement can be improved.

This invention provides a quantum-derived Newton-Raphson optimal fractional-order spectrogram generation method, comprising the following steps: Step 1: Acquire the original audio signal and construct a fractional-order spectrogram based on fractional Fourier transform (FRFT); Step 2: Perform nonlinear scaling compression on the fractional-order spectrogram using a Mel filter bank to generate a fractional-order Mel spectrogram; Step 3: Construct an adaptive optimization framework with information entropy minimization as the objective function to measure the information fidelity between the spectrogram and the original signal; Step 4: Use the quantum-derived Newton-Raphson optimization algorithm (QNRBO) to globally optimize the fractional-order order, frame length, and frame shift hyperparameters to generate the optimal fractional-order spectrogram; Step 5: Input the optimal fractional-order spectrogram into a downstream speech recognition model. This technical solution aims to systematically solve core problems such as insufficient traditional time-frequency representation capabilities, rigid hyperparameter configuration, limited optimization algorithm performance, and feature-task disconnect.

The present disclosure provides a physical information neural network-based pose estimation method, device and equipment, relating to the technical field of computer, the method comprising: collecting an original sensor sequence and preprocessing the original sensor sequence to obtain a time-frequency representation; determining a multi-scale feature and a residual attention weight based on the time-frequency representation; inputting the multi-scale feature and the residual attention weight into a physical information neural network model to obtain an angular velocity estimation and a pose quaternion update term; calculating a physical residual loss according to the angular velocity estimation and the pose quaternion update term and a physical kinematics constraint, and constructing a total loss function based on the physical residual loss and a measurement consistency loss; optimizing network parameters of the physical information neural network model based on the total loss function to obtain a trained pose estimation model; compressing and deploying the pose estimation model, and generating a pose estimation result based on real-time collected sensor data through the pose estimation model. Low-cost and high-precision pose estimation can be achieved.

The invention relates to the technical field of artificial intelligence and machine learning, and discloses a method and a system for realizing broadband noise point identification of an automobile audio based on time-frequency analysis, and the method comprises the steps: adding a variable-length window function of an audio analysis fragment, obtaining a windowed audio analysis fragment, analyzing an initial time-frequency spectrum of the windowed audio analysis fragment, and obtaining an initial time-frequency spectrum of the windowed audio analysis fragment; generating an instantaneous frequency estimation matrix of the windowed audio analysis segment; combining the instantaneous frequency estimation matrix with the initial time-frequency spectrum to obtain enhanced time-frequency representation of the windowed audio analysis fragment, analyzing noise time-frequency units of the enhanced time-frequency representation, and communicating the noise time-frequency units to obtain a broadband noise point candidate region of the windowed audio analysis fragment; and analyzing the real noise point probability of the broadband noise point candidate region, and determining broadband real noise points of the broadband noise point candidate region. According to the invention, the efficiency and accuracy of automobile audio broadband noise point identification can be improved.

The invention discloses a keyless phase-change rotating speed bearing fault diagnosis method and device, and belongs to the technical field of bearing fault diagnosis. The method comprises the following steps: acquiring a variable rotating speed bearing fault vibration signal; a pre-constructed TFRPET algorithm is applied to the vibration signals, and high-resolution time-frequency representation TFR is obtained; the TFRPET algorithm compresses the width of a TF ridge through a peak value extraction operator PEO, and the energy aggregation of the TF ridge is improved; detecting a TF ridge from a time-frequency representation (TFR) according to a ridge detection technique and selecting a reference instantaneous frequency (IF) from the TF ridge; and carrying out angular domain resampling and order spectrum diagnosis on the vibration signal by using the reference instantaneous frequency IF to obtain a fault diagnosis result of the keyless phase change rotating speed bearing. The method does not need to depend on a key phase device, can adapt to a variable rotating speed working condition, achieves high-resolution TFR through the TFRPET algorithm, and solves the problem that in the prior art, the TFR resolution is insufficient, and the bearing fault judgment accuracy is affected.

The application discloses a kind of water turbine cavitation degree detection methods based on cavitation perception multi-path deep learning network model, comprising: obtaining acoustic signal sequence collected from water turbine equipment and normalizing, construct including one cavitation feature extraction path, one periodicity feature extraction path, one multi-scale mel spectrum extraction path of extracting time-frequency representation feature, feature fusion module, main classifier and auxiliary classifier in the cavitation perception multi-path deep learning network model, feature fusion module is adaptively fused to the cavitation feature and periodicity feature extracted, and learns fusion weight;Again, the physical feature vector after fusion is input into auxiliary classifier and classified, and multi-scale mel spectrum feature is input into main classifier and classified;And based on including main classification loss, auxiliary classification loss and cavitation perception alignment loss in the target cost function, the model is trained.The application can provide interpretable physical feature representation while maintaining high classification accuracy.

According to embodiments of this disclosure, a method, apparatus, device, and storage medium for text-to-speech conversion are provided. The method includes generating a predicted speech representation of the target text read by a first speaker, based on a target text to be converted and a first timbre of a first speaker. The predicted speech representation indicates speech features that vary over time. The method further includes generating a predicted time-frequency representation of the target text read by a second speaker, based on the predicted speech representation and a second timbre of a second speaker. The predicted time-frequency representation indicates the speech signal strength that varies over time at different frequencies. The method further includes converting the predicted time-frequency representation into audio of the target text read by the second speaker. This reduces the difficulty of prediction and improves the sound quality of the generated audio.

The invention discloses an automatic detection method for an uplink of a ground measurement and control station, and relates to the technical field of aerospace measurement and control and satellite communication operation guarantee. The separability of the uplink signal in the aliasing environment is obviously improved; candidate components are obtained based on blind source separation with mutual information minimization as a target, modulation identification is carried out through a depth model driven by time-frequency representation, and robust distinguishing of different satellite service systems is achieved. The carrier frequency, the symbol rate, the modulation category and the task parameter set are subjected to weighted similarity matching in a tolerance range, and one-to-one constraint solution is carried out, so that mismatching and misjudgment caused by near-frequency, near-speed or fast switching can be effectively inhibited; when observation environment or orbit parameters change, separation and window configuration can be adaptively adjusted, and real-time performance and accuracy are maintained.

The invention discloses a bearing fault diagnosis method based on boundary rearrangement Chirplet transformation. The method comprises the following steps: S1, collecting a bearing fault signal and a rotating speed signal by using an acceleration sensor and an encoder; s2, generalized matching Chirplet transformation is constructed to process the bearing fault vibration signal, and a corresponding time-frequency representation result is obtained; s3, searching a local amplitude maximum value represented by time frequency along the time direction, and defining a new time rearrangement strategy; s4, designing a compression boundary algorithm based on the energy distribution characteristics to reduce the influence of environmental noise; s5, energy in the compression boundary is redistributed according to a time rearrangement strategy, and the time interval of two adjacent fault impacts extracted is calculated based on the obtained time frequency representation result; and S6, comparing the time interval with the actually extracted time interval of the fault impact, and further judging the fault occurrence position of the bearing. According to the method, the energy aggregation of time-frequency representation is improved, the noise robustness is high, and a powerful guarantee is provided for extraction of fault features of the rotating machine bearing.

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 application discloses a kind of sparse generalized W transform gas reservoir characterization time-frequency analysis method, for the first time L1 norm is introduced to the mathematical relationship between generalized W transform and seismic signal for sparse constraint, and is solved using Bregman iteration algorithm, to obtain a kind of analysis result with higher time-frequency resolution.The method absorbs the advantage that generalized W transform highlights low-frequency information of seismic signal, avoids the problem of main frequency splitting, and provides a more sparse time-frequency representation for non-stationary seismic signals. When applied to the time-frequency analysis of actual seismic data, it can provide a high-precision seismic spectral decomposition result for gas reservoirs, thereby more accurately delineating the low-frequency abnormal area of gas reservoir.

The invention provides a bearing fault diagnosis method and system based on sample enhancement and domain self-adaption, and belongs to the field of industrial intelligent manufacturing, and the method comprises the steps: employing wavelet transform, converting an original vibration signal into time-frequency representation, taking the time-frequency representation as an image and a time sequence, extracting image features and time sequence features, and carrying out the recognition of the image features and the time sequence features; carrying out feature fusion on the image features and the time sequence features through a weight network; inputting the fused features into a classification network to obtain a classification result of fault diagnosis; constructing an enhanced sample as a new time-frequency representation according to the size of the first attention score; a plurality of bearing data sets disclosed in other fields are used as a source domain, collected original vibration signals of the bearing are used as a target domain, and further training is performed on an image encoder, a time sequence encoder and a gating mechanism based on the source domain and the target domain. According to the method, efficient and accurate bearing fault diagnosis is realized under the condition of limited samples.

The application provides a robust denoising processing method and system enhanced by acoustic wave signal self-supervised learning, relates to the technical field of signal processing, and comprises the following steps: performing feature enhancement on an initial time-frequency representation through a dynamic self-adaptive mask strategy; constructing a self-supervised reconstruction task based on the mask time-frequency representation; separating noise and signal subspaces in a semantic manifold space; combining time-domain continuity features to establish a probability propagation network to model local dependent relationships; and finally generating denoising weights and realizing semantic fidelity optimization. The application can effectively improve the denoising effect and semantic integrity of acoustic wave signals in a noise complex environment.

A method includes receiving an audio input and generating a noisy time-frequency representation based on the audio input. The method also includes providing the noisy time-frequency representation to a noise management model trained to predict a denoising mask and a signal presence probability (SPP) map indicating a likelihood of a presence of speech. The method further includes determining an enhanced spectrogram using the denoising mask and the noisy time-frequency representation. The method also includes providing the enhanced spectrogram and the SPP map as inputs to a keyword classification model trained to determine a likelihood of a keyword being present in the audio input. In addition, the method includes, responsive to determining that a keyword is in the audio input, transmitting the audio input to a downstream application associated with the keyword.

The application discloses a radar human behavior recognition method based on multi-domain feature fusion, and is applied to the problem that the existing human behavior recognition method based on a radar sensor and a deep learning technology only adopts the features of one domain or only adopts one time-frequency analysis method in a time-frequency domain, thereby causing insufficient expression of human behavior features; three time-frequency analysis methods with different time-frequency resolutions, namely, a short-time Fourier transform, an adaptive optimal kernel time-frequency representation method and a Hann kernel reduction cross-term interference distribution, are selected; three types of frequency spectrum diagrams are obtained; then, the three types of time-frequency spectrum diagrams are combined and used by using a SE Net and a 3DCNN network in the time-frequency domain, so that human behavior features are more fully expressed; in a distance domain, a key feature is extracted by using the SE Net; and the features of two domains are combined, the mutual relationship is found, and the recognition accuracy of human behaviors is effectively improved.

The application discloses a radar signal intra-pulse modulation recognition method based on LMSST and DCNN. The application estimates the instantaneous frequency based on a ridge extraction algorithm by adopting an LMSST time-frequency distribution, and detects the local maximum value of a time-frequency spectrum in a frequency direction, so that the energy aggregation is improved, and the time-frequency representation characteristics under different modulations are significantly improved. Meanwhile, the deep cascaded convolutional neural network improves the multi-scale resolution of the convolutional neural network, can solve the problems of weak learning ability, low generalization ability and poor clustering effect of the deep model, so that the accuracy of LPI radar signal modulation recognition is improved, automatic classification of radar modulation waveforms is realized, the workload of electronic reconnaissance professionals is reduced, and subsequent situation and threat estimation is facilitated.

The present disclosure provides a method for audio object estimation, comprising: obtaining a time-frequency representation of an audio signal that comprises a mixture of at least a first signal and a second signal of different types; obtaining a set of bin features from the time-frequency representation, wherein each bin feature in the set corresponds to a respective time-frequency bin of the time-frequency representation; feeding the set of bin features to a neural network model to estimate a first mask and a second mask respectively corresponding to the first signal and the second signal; and respectively applying the first mask and the second mask to the time-frequency representation, in order to obtain an estimated time-frequency representation of the first signal and an estimated time-frequency representation of the second signal.

The invention discloses a non-stationary signal time-frequency analysis method based on an adaptive offset window. The method comprises the following steps: preprocessing an input signal to obtain an analysis signal; calculating instantaneous frequency and identifying trend characteristics of the instantaneous frequency; constructing a self-adaptive offset window according to the instantaneous frequency trend characteristics; the method comprises the following steps: constructing self-adaptive bias Chirplet transformation by fusing a self-adaptive bias window and generalized linear Chirplet transformation; and a corresponding time-frequency spectrum is obtained by using adaptive bias Chirplet transformation. According to the method, the energy diffusion phenomenon under dense frequency components can be effectively inhibited, the aggregation and resolution of time-frequency representation are remarkably improved, accurate description of instantaneous characteristics is realized, and excellent anti-noise performance is shown.

This invention relates to the field of power system fault signal processing, specifically to a power fault time-frequency analysis method based on synchronization extraction transform. The invention employs a Rogowski coil sensing system to acquire transient fault signals, obtains an initial time-frequency representation through short-time Fourier transform, and derives a corrected time-frequency representation using the derivative of a window function. Based on the ratio of the two representations, the instantaneous frequency trajectory is calculated. A synchronization extraction operator is constructed using a binarization criterion with discrete frequency intervals as an adaptive threshold. Only time-frequency coefficients on the instantaneous frequency trajectory are retained, rather than all coefficients in the concentrated energy diffusion region, resulting in a highly concentrated energy time-frequency representation. This method significantly improves time-frequency resolution while exhibiting stronger noise robustness and multi-component signal separation capabilities, providing an effective technical means for rapid and accurate fault location in distribution networks.

The invention relates to the technical field of time-frequency analysis and modal decomposition, and discloses a time-frequency modal decomposition method and system based on deep learning, a terminal and a storage medium, and the method comprises the steps: obtaining an input complex number of time-domain signals, analyzing the complex number of time-domain signals through a high-resolution time-frequency analysis module of a convolutional neural network model, and obtaining a complex number of time-domain signals; obtaining multi-scale time-frequency representation; performing instance segmentation on the multi-scale time-frequency representation through a time-frequency mode segmentation module of a convolutional neural network model to obtain a time-frequency mode; and performing trajectory prediction on the time-frequency mode and the complex number time-domain signal through a signal reconstruction module of a convolutional neural network model to obtain a signal component. According to high-resolution time-frequency analysis and time-frequency modal segmentation, separation and trajectory prediction are carried out on aliasing components in complex time-domain signals, and accurate extraction of signal components is realized.

A converter valve recording data extraction method based on adaptive filtering and data driving belongs to the technical field of converter valve fault identification, and comprises the following steps: recording data of a converter valve under different working conditions are extracted and processed, and the recording data comprise core variable parameters under each converter valve fault type; based on the converter valve fault types, constructing a core variable parameter sensitivity analysis and calculation model, extracting the core variable parameter sensitivity degree under each converter valve fault type, and then performing sorting according to the core variable parameter sensitivity degree; a 3D-CNN three-dimensional convolutional neural network is constructed, corresponding three-dimensional time-frequency representation is constructed through a core variable parameter sensitive degree sorting result to serve as network input, then a self-attention module is introduced to highlight key features, and finally a corresponding high-dimensional feature vector capable of accurately representing the fault type is extracted from a network middle layer. Through improvement of an algorithm, accurate and rapid extraction of converter valve fault recording data characteristics is realized.

This invention provides a system and method for estimating and suppressing spatial noise using machine learning support. [Solution] The noise estimation and suppression method includes the steps of: estimating the probability of speech and noise for each band of the input audio signal using a machine learning classifier; estimating a set of averages of speech and noise or a set of averages and covariances of speech and noise based on the probability and microphone covariance over the bands using a directional model; estimating the average and covariance of noise power based on the probability and power spectrum using a level model; determining a first noise suppression gain based on the directional model; determining a second noise suppression gain based on the level model; selecting the first noise suppression gain or the second noise suppression gain or their sum based on the signal-to-noise ratio of the input audio signal; and scaling the time-frequency representation of the input signal by the selected noise suppression gain.

Methods and systems of processing audio data with a multi-stage audio front end model is provided. A one-dimensional audio waveform is received as input and processed using a multi-stage audio frontend model to convert the one-dimensional waveform into a two-dimensional matrix representing features of the audio waveform. The multi-stage learnable audio frontend model is configured to apply a first filterbank to the audio waveform to generate a first time-frequency representation of the audio waveform; apply a first decimation filter to the audio waveform to generate a first decimated audio input; apply a second filterbank to the first decimated audio input to generate a second time-frequency representation of the audio waveform; and stack the first time-frequency representation and the second time-frequency representation together to generate the two-dimensional matrix.

Sound reconstruction from corresponding auditory neurograms was a long-standing problem in computational neurosciences and auditory modeling fields, and was solved by my end-to-end artificial neural network models. This is a novel system for sound synthesis utilizes the modern neural vocoders and spiking-based time-frequency representations derived from auditory neurograms. This method introduces the auditory neurograms as the novel input, replacing the mel-spectrogram in conventional vocoding approaches. This method provides high-fidelity sound quality of synthesized sound, compared to the existing two-stage models on the sound reconstruction task and the modern neural vocoders. This method also provide zero-shot performance on unseen datasets. This method can be used for the applications of a novel Sound Synthesizer using auditory spiking data and Speech NeuroProsthesis using auditory spiking activities.

The invention relates to an oscillation signal detection method and device, equipment, a storage medium and a program product. The method comprises the following steps: acquiring a time domain voltage signal from a power system; performing time-frequency analysis on the time-domain voltage signal based on time-frequency transformation to obtain a first time-frequency representation complex matrix; time-frequency rearrangement is carried out on the first time-frequency representation complex matrix to obtain a second time-frequency representation complex matrix, the energy concentration ratio of time-frequency representation of the second time-frequency representation complex matrix is higher than that of the first time-frequency representation complex matrix, and a time-frequency representation spectrogram is generated based on the second time-frequency representation complex matrix; extracting an instantaneous frequency spectral line from the time-frequency representation spectrogram, and carrying out instantaneous frequency identification to obtain an instantaneous frequency value; and based on the instantaneous frequency value, determining whether harmonic waves and / or inter-harmonic waves exist in the time domain voltage signal, and in response to the existence of the harmonic waves and / or inter-harmonic waves in the time domain voltage signal, determining that an oscillation signal exists in the power system. The oscillation signal in the power system can be accurately detected, and a basis is provided for stability analysis of the power system.

This invention provides a method and system for training an artificial neural network model to extract features indicating variations in psychoacoustic attributes from digital audio signals and generate contextual latent spatial representations. [Solution] The learning system consists of a read / decode logic that reads a specific digital audio signal source from a set of digital audio signal sources associated with a specific sound content category, a transform logic that generates a time-frequency representation based on the specific digital audio signal, and a learning logic that uses an artificial neural network to learn a set of numerical codes that provide a latent spatial representation of the time-frequency representation. The read, generate, and learn process is repeated to train the artificial neural network. The model database retrieves a set of model parameters learned from the trained artificial neural network and stores them in a computer storage medium.

The invention discloses a single-wavelength airborne sounding laser radar sea-land waveform classification method, which belongs to the technical field of remote sensing detection and signal processing, is used for sea-land waveform classification, and comprises the following steps: obtaining original waveform data, and carrying out preprocessing by adopting a wavelet soft threshold and continuous wavelet transform to obtain a one-dimensional echo waveform effective signal and a two-dimensional time-frequency graph; and a dual-path multi-modal feature fusion network is constructed, the one-dimensional echo waveform effective signal and the two-dimensional time-frequency graph are used as input, a sea-land waveform classification result is output, and the dual-path multi-modal feature fusion network comprises an attention convolution residual network module, a one-dimensional convolution neural network time feature extraction module and a multi-modal data fusion module. According to the method, the time-frequency representation is introduced to enhance the discrimination information, the feature content is enriched, the two types of complementary information of the time-domain waveform and the time-frequency representation are utilized to construct a joint feature learning and fusion mechanism, the waveform confusion in a single feature space is reduced, and the classification stability of the near-shore land and water boundary region is improved.

The invention relates to a broadband oscillation signal detection method and device, equipment and a storage medium. The method comprises the following steps: performing short-time Fourier transform on an original time domain signal to obtain a short-time Fourier transform complex matrix; the short-time Fourier transform complex matrix is sharpened through Fourier synchronous compression transform to obtain a sharpened time-frequency representation spectrogram, the energy diffusion range in the time-frequency representation spectrogram is compressed, and the frequency resolution is improved; extracting a plurality of instantaneous frequency spectral lines from the sharpened time-frequency representation spectrogram according to a penalty optimization algorithm; for each instantaneous frequency spectral line, determining a frequency value corresponding to the instantaneous frequency spectral line; according to the multiple relation between the frequency value and the preset power frequency, the type of the instantaneous frequency spectral line is judged, accurate detection of the type of the instantaneous frequency spectral line is achieved, and safe and stable operation of a power system is guaranteed.