38 results about "Spectral representation" patented technology

The invention discloses an intelligent identification method and system for a dishwashing detergent volatile matter spectrogram, and relates to the technical field of spectrogram characterization learning and intelligent identification, and the method comprises the steps: setting a temperature time sampling grid, and collecting and constructing a dishwashing detergent volatile matter spectrogram tensor. A spectrogram view is generated based on the physical reasonable enhancement operator set, and spectrogram tensor self-supervised contrast learning is carried out to obtain a spectrogram representation vector. And fitting the volatilization kinetic parameters according to the spectrogram representation vector, fusing the volatilization kinetic parameters and outputting an identification result. Static spectrogram recognition of the tableware detergent volatile matter is upgraded to temperature-time dynamic modeling, temperature-time tensor data capable of being aligned is provided, robust characterization is obtained through self-supervised learning under physical consistency enhancement constraint, dynamic fingerprints are extracted and subjected to fusion recognition with characterization, and a recheck closed loop is provided. Therefore, an enforceable and reviewable identification process is still kept under the conditions of drifting, noise, missing spectrum segments and small samples.

An audio device is disclosed, comprising multiple microphones and an audio module. The multiple microphones generate multiple audio signals. The audio module coupled to the multiple microphones comprises a processor, a storage media and a post-processing circuit. The storage media includes instructions operable to be executed by the processor to perform operations comprising: producing multiple instantaneous relative transfer functions (IRTFs) using a known adaptive algorithm according to multiple spectral representations for multiple first sample values in current frames of the multiple audio signals; and, performing distractor suppression over the multiple spectral representations and the multiple IRTFs using an end-to-end neural network to generate a compensation mask. The post-processing circuit generates an audio output signal according to the compensation mask. Each IRTF represents a difference in sound propagation between each predefined microphone and a reference microphone of the microphones relative to sound sources.

The invention relates to a method and system for reconstructing a hyperspectral image based on an RGB image, and the method comprises the steps: S1, selecting an RGB image in a space; s2, processing the RGB image to obtain end member information of the RGB image; s3, generating end member spectral representation based on the end member information of the RGB image; s4, encoding, decoding and attention modeling are carried out on the RGB image, and abundance information of the RGB image is obtained; and S5, constructing a rough hyperspectral image based on the end member information of the RGB image and the abundance information of the RGB image, performing feature enhancement on the rough hyperspectral image and the RGB image, and performing fusion to obtain a final hyperspectral image. According to the invention, the hyperspectral image can be effectively constructed based on the RGB image.

This invention proposes a method and system for ECG clustering based on multi-lead semantic consistency, belonging to the field of ECG signal processing technology. The method includes: constructing a similarity correlation matrix for an ECG signal set using adaptive graph learning; learning the spectral representation of each lead of the ECG signal using spectral clustering on the similarity correlation matrix, whereby spectral clustering utilizes multi-lead shared spectral embedding to measure the correlation between ECG signals, obtained through an optimized minimum edge weight objective function; wherein the optimized minimum edge weight objective function includes the multi-lead shared semantic similarity matrix; after obtaining the spectral embedding matrix, constructing a spectral rotation objective function by introducing an orthogonal rotation factor matrix; merging the minimum edge weight objective function and the spectral rotation objective function into a total objective function, and optimizing the total objective function using the alternating direction multiplier method to obtain the ECG clustering result. This invention significantly improves the clustering quality through the learning of semantically consistent graphs.

The invention discloses a spectral band selection method based on fractional domain transformation and differentiable information imbalance. The spectral band selection method is used for obtaining a band subset with diagnosis information. The method comprises the following steps: S1, collecting a medical hyperspectral image, and dividing the medical hyperspectral image into a training-verification set and a test set; s2, performing fractional Fourier transform on the hyperspectral data of the training-verification set to obtain transform domain spectral representations under different orders; s3, evaluating the retention degree of the candidate wave band subset on the transform domain spectral characteristics under different orders based on the differentiable information imbalance, and carrying out wave band selection; and S4, determining an optimal transformation order, performing fractional Fourier transform and waveband selection processes of the test set data, and generating an optimal waveband subset. According to the method, information most related to lesion judgment can be effectively reserved while the data dimension is reduced, the recognition accuracy and robustness of the lesion area in the medical hyperspectral image are effectively improved, and the method has important clinical application value.

The application discloses a kind of progressive space-time spectrum characterization learning cultivated crop type monitoring systems, comprising;Remote sensing image acquisition module for obtaining the multi-temporal multispectral remote sensing image of target area on covering cultivated crop complete growing season;NDVI extraction module for calculating NDVI according to each phase ground surface real reflectivity time series image data;Spectrum perception module for reconstructing the time series spectral features of cultivated crop;Time series perception module for obtaining the time series aggregation features of cultivated crop based on the phenology curve of cultivated crop to the reconstructed time series spectral features of cultivated crop;Spatial modeling module for introducing position coding on the time series aggregation features of cultivated crop based on the field location of cultivated crop, spatial modeling is carried out, and the spatial features of cultivated crop are obtained;Classification module for generating the type distribution map of cultivated crop according to cultivated crop spatial features.

An apparatus for generating a synthesis audio signal using a patching control signal has a first converter, a spectral domain patch generator, a high frequency reconstruction manipulator and a combiner. The first converter is configured for converting a time portion of an audio signal into a spectral representation. The spectral domain patch generator is configured for performing a plurality of different spectral domain patching algorithms, wherein each patching algorithm generates a modified spectral representation having spectral components in an upper frequency band derived from corresponding spectral components in a core frequency band of the audio signal. The spectral domain patch generator is furthermore configured to select a first spectral domain patching algorithm from the plurality of patching algorithms for a first time portion and a second spectral domain patching algorithm from the plurality of patching algorithm for a second different time portion in accordance with the patching control signal to obtain the modified spectral representation.

A computer-implemented method for generating a spectral model representation of an electromagnetic element for determining an electromagnetic response to electromagnetic radiation interacting with the electromagnetic element is presented. The spectral representation contains a two-dimensional array of elements corresponding to respective wave vectors, each of which is defined by (different) respective pairs of wave numbers in two directions transverse to each other. The electromagnetic elements are geometrically described as a plurality of slices stacked transversely to these directions, and each slice contains one or more polygons. The method includes determining points along each of the polygon edges, a calculation coefficient being calculated based on a range of the corresponding polygon edge in a lateral direction. By performing a fast Fourier transform (FFT) algorithm using the coefficients, Fourier components are generated in each of the lateral directions for each of the wave vectors.

The invention relates to a corn kernel quality evaluation method, a machine readable storage medium and equipment, and belongs to corn kernel quality evaluation in the field of agricultural products, the method comprises the following steps: firstly obtaining the absorbance spectrum and the protein content of each sample in a sample set, then preprocessing the absorbance spectrum, converting a spectrum view into a graph structure in a graph neural network, constructing a quality evaluation model, sequentially training and evaluating the model, and finally evaluating the quality of the corn kernels by using the established quality evaluation model. According to the method, explicit modeling and feature fusion are carried out on potential relationships among all views, and adaptive enhancement of key spectral information and suppression of redundant features are realized; by combining hierarchical feature extraction and a multi-scale fusion strategy, the obtained spectral representation has both a global trend and local details, and high prediction precision and good robustness can still be kept under the condition of large sample difference or limited data scale.

The invention provides an ocean flow field reconstruction method based on multi-modal hybrid filtering and latent code modulation, and relates to the technical field of ocean dynamics analysis, and the method comprises the steps: carrying out the data preprocessing, constructing a space-time sample index, and obtaining a space sampling point coordinate; carrying out context embedding, carrying out mixed frequency coding, then carrying out multiplicative interaction and hierarchical condition modulation to form frequency selective feature representation, and carrying out deep residual stabilization to obtain implicit features of a current residual layer; the operation of the multiple residual layers is repeated until final implicit features are obtained; and generating a reconstructed flow field through coordinate-by-coordinate linear projection based on the final implicit features. According to the invention, tunable multi-scale spectrum expression is formed through mixed frequency coding. And large-scale flow state characteristics and environment background information are represented through hierarchical condition modulation based on context embedding. A multiplicative conditional modulation mechanism is constructed to strengthen high-frequency response, and layer normalization, learnable layer scaling and a random depth regularization mechanism are introduced in a deep feature propagation stage to carry out deep residual stabilization.

A system and method are provided for monitoring a factory, comprising during a training period of time, training a machine learning model with training data including a plurality of known spectral representations each of which is designated in the training data as normal or abnormal; and during an operating period of time receiving a first audio sample from a first microphone in an array of microphones suspended above a plurality of mechanical devices, generating a first spectral representation from the first audio sample, providing the first spectral representation to the machine learning model, and determining that the first spectral representation is abnormal.

This application discloses a training method and apparatus for an audio codec model. The main technical solution includes: acquiring training data comprising multiple audio samples; training the audio codec model using the training data; wherein the training includes: extracting the spectral representation of the audio samples; extracting the acoustic feature representation of the audio samples using a first audio encoder; mapping the acoustic feature representation to a latent space using a second audio encoder to obtain the latent space feature representation of the audio samples; masking the latent space feature representation to obtain the masked latent space feature representation; performing decoding processing based on the latent space feature representation using an audio decoder to obtain the reconstructed spectrum; and updating the model parameters of the audio codec model using the reconstructed spectrum and the spectral representation. This approach can improve the semantic understanding capability of the audio codec model, thereby improving the performance of speech generation tasks.

An apparatus for generating a synthesis audio signal using a patching control signal has a first converter, a spectral domain patch generator, a high frequency reconstruction manipulator and a combiner. The first converter is configured for converting a time portion of an audio signal into a spectral representation. The spectral domain patch generator is configured for performing a plurality of different spectral domain patching algorithms, wherein each patching algorithm generates a modified spectral representation having spectral components in an upper frequency band derived from corresponding spectral components in a core frequency band of the audio signal. The spectral domain patch generator is furthermore configured to select a first spectral domain patching algorithm from the plurality of patching algorithms for a first time portion and a second spectral domain patching algorithm from the plurality of patching algorithm for a second different time portion in accordance with the patching control signal to obtain the modified spectral representation.

An apparatus for generating a synthesis audio signal using a patching control signal has a first converter, a spectral domain patch generator, a high frequency reconstruction manipulator and a combiner. The first converter is configured for converting a time portion of an audio signal into a spectral representation. The spectral domain patch generator is configured for performing a plurality of different spectral domain patching algorithms, wherein each patching algorithm generates a modified spectral representation having spectral components in an upper frequency band derived from corresponding spectral components in a core frequency band of the audio signal. The spectral domain patch generator is furthermore configured to select a first spectral domain patching algorithm from the plurality of patching algorithms for a first time portion and a second spectral domain patching algorithm from the plurality of patching algorithm for a second different time portion in accordance with the patching control signal to obtain the modified spectral representation.

A system and method for enhancing a call characterization of voice calls is disclosed. The method comprises: receiving, at the platform (1), the voice call (30) segmented in audio packets; generating a Call Detail Record "CDR" (12) comprising call parameters; processing, by the Digital Signal Processing "DSP" engine (300), the audio packets, wherein the processing step in turn comprises: computing audio spectral representations; and, accumulating a predetermined number of audio packets (30.1). The method further comprises: evaluating, by the Audio Segmentation "AS" engine (301), an audio type present in the accumulated audio packets; generating temporal marks (31), by the Audio Segmentation "AS" engine (301), indicating which audio segment correspond to each audio type; computing metrics (32), by the Metrics Computation "MC" engine (302), using the accumulated audio packets and the temporal marks; generating an enhanced Call Detail Record "eCDR" (33) by adding the metrics (32) to the "CDR".

Systems, methods, and non-transitory computer readable medium are disclosed for encoding tissue behavior. A bioreactor may be implemented, where the bioreactor comprises a device configured for growing tissue and a sensor assembly configured to detect functional response(s) of a tissue within the device. The systems and methods may also include a control unit communicatively coupled to the bioreactor and configured to obtain, from the bioreactor, waveform(s) comprising functional response(s) of the tissue within the device over a time frame. The control unit may also be configured to transform the one or more waveforms to spectral representation(s) that provide frequency-based characterizations of the functional response(s) of the tissue during the time frame thereby encoding a behavior of the tissue.

Method and system for electric vehicle charging The electric charging method is implemented by a system comprising a supervisory module, a plurality of charging stations and a compensation device adapted to inject a compensation current into the distribution network, each charging station being configured to communicate with the supervisory module via a communication link, the method comprising obtaining, from each charging station, a spectral representation of the current consumed by the station, a determination (72) of an overall spectral representation of the charging system from the spectral representations of the current consumed per charging station and per spectral rank, a calculation (75) of at least one parameter of at least one compensation current, as a function of the overall spectral representation and at least one predetermined power factor degradation threshold.The compensation current is injected into the electricity distribution network. See Figure 4 for the abbreviation.

The invention provides a self-supervised implicit hyperspectral mixed noise removal method coupled with low-rank prior, and relates to the field of hyperspectral remote sensing image denoising, the method comprises the following steps: realizing space and spectrum decoupling through low-rank characteristics, and designing an implicit space mapping network and an implicit spectrum representation network; respectively mapping input two-dimensional space coordinates and one-dimensional spectral coordinates to a high-dimensional feature space by using a coordinate position encoder; an image feature encoder excavates deep features of spatial and spectral dimensions of the hyperspectral image; after being spliced, the high-dimensional coordinate features and the image features are projected to a target signal space through a multi-layer perceptron decoder, and a restored low-rank subspace is multiplied by a spectrum substrate to obtain a denoised hyperspectral image. According to the method, a degraded image is taken as a supervision signal, an implicit neural representation self-supervision framework is constructed, total variation regularization constraint is introduced, violent fluctuation of local pixels is inhibited by using image continuity, multi-type high-intensity mixed noise is effectively removed, and image quality and availability are remarkably improved.

A method, computer program product, and computing system for enhancement of audio signals received from a plurality of microphones. A multichannel audio signal is received from a plurality of microphones and is processed with a short-time discrete cosine transform (STDCT) to generate a real-valued spectral representation of the multichannel signal encoding both magnitude and phase information. Magnitude- and phase-dependent weights are generated, and an enhanced single-channel signal is produced based upon, at least in part, the spectral representation of the multichannel signal and the magnitude- and phase-dependent weights.

An apparatus for generating a synthesis audio signal using a patching control signal has a first converter, a spectral domain patch generator, a high frequency reconstruction manipulator and a combiner. The first converter is configured for converting a time portion of an audio signal into a spectral representation. The spectral domain patch generator is configured for performing a plurality of different spectral domain patching algorithms, wherein each patching algorithm generates a modified spectral representation having spectral components in an upper frequency band derived from corresponding spectral components in a core frequency band of the audio signal. The spectral domain patch generator is furthermore configured to select a first spectral domain patching algorithm from the plurality of patching algorithms for a first time portion and a second spectral domain patching algorithm from the plurality of patching algorithm for a second different time portion in accordance with the patching control signal to obtain the modified spectral representation.

The present application relates to a kind of distributed acoustic sensing privacy protection methods based on learnable feature decoupling, comprising: obtaining the original signal containing multiple spatial channels collected by distributed acoustic sensing system;The original signal is mixed to obtain enhanced signal;Each spatial channel of enhanced signal is carried out short-time Fourier transform and phase gating is introduced, to obtain adaptive spectral representation;Adaptive spectral representation is arranged as spatial sequence, while introducing position mixed gating, and the feature of each spatial channel enhanced by position is output by Transform encoder;Introduce pooling gating, and after each spatial channel feature is pooled, privacy protection representation is obtained by fusion, finally, the strategy of alternating gradient descent is used to optimize adversarial learning network.The method can adaptively decouple task-related features and privacy-sensitive attributes in DAS signal, and automatically balance privacy protection and task utility under the adversarial learning framework, to realize the optimal balance of privacy protection and task utility.

This invention discloses a method and system for identifying the authenticity of Pinellia ternata and its processed products based on hyperspectral imaging and transfer learning networks. The method involves acquiring hyperspectral image data of Pinellia ternata and its processed products, performing black-and-white correction on the hyperspectral image data, and extracting the spectral sequence from the source domain data. An S-STN transfer learning framework is constructed and trained using the spectral sequence from the source domain data. The S-STN transfer learning framework includes a spectral encoder and a capsule classifier. A masked spectral reconstruction strategy is used to perform self-supervised pre-training of the spectral encoder, learning device-independent general spectral representations. The parameters of the pre-trained spectral encoder are frozen, and the capsule classifier is trained in a supervised manner on the source domain task to obtain a preliminary identification model. This preliminary identification model is then transferred to the target domain task, the spectral encoder is frozen, and only the capsule classifier is fine-tuned to output the authenticity identification result of the target domain sample.

A device and method for manipulating an audio signal includes a windower for generating a plurality of consecutive blocks of audio samples, the plurality of consecutive blocks including at least one padded block of audio samples, the padded block having padded values and audio signal values. a first converter for converting the padded block into a spectral representation having spectral values, a phase modifier for modifying phases of the spectral values to obtain a modified spectral representation and a second converter for converting the modified spectral representation into a modified time domain audio signal.

The invention discloses a random seismic oscillation dimensionality reduction modeling method of a non-iterative high-matching standard response spectrum, and belongs to the technical field of engineering structure earthquake resistance, and the method comprises the steps: S1, building a stable seismic oscillation power spectrum model; s2, establishing a non-stationary seismic oscillation evolution power spectrum model; s3, expressing a non-stationary seismic oscillation process as a function form of a single random variable by adopting a dimension reduction method based on a spectral representation-random orthogonal function; s4, determining undetermined parameters of the evolution power spectrum model based on a particle swarm optimization algorithm by taking a standard response spectrum as a target; and S5, generating a representative sample set of the seismic oscillation acceleration, calculating an average response spectrum, and comparing the average response spectrum with a standard response spectrum. According to the method, the average response spectrum and the standard response spectrum of the random seismic oscillation representative sample set generated at a time can be directly fitted with high precision, and iteration is not needed; and the representative sample set generated by the method is a complete probability set, and a foundation is laid for realizing anti-seismic reliability analysis of a complex engineering structure in a total probability sense.

An apparatus for generating a synthesis audio signal using a patching control signal has a first converter, a spectral domain patch generator, a high frequency reconstruction manipulator and a combiner. The first converter is configured for converting a time portion of an audio signal into a spectral representation. The spectral domain patch generator is configured for performing a plurality of different spectral domain patching algorithms, wherein each patching algorithm generates a modified spectral representation having spectral components in an upper frequency band derived from corresponding spectral components in a core frequency band of the audio signal. The spectral domain patch generator is furthermore configured to select a first spectral domain patching algorithm from the plurality of patching algorithms for a first time portion and a second spectral domain patching algorithm from the plurality of patching algorithm for a second different time portion in accordance with the patching control signal to obtain the modified spectral representation.

The present invention relates to an imaging device (100), in particular an endoscopic imaging device, comprising: an image capturing unit (102), which is designed to map an object region and has a first capturing strand (104) which is designed to capture reference images (106) of the object region, and a second capturing strand (108) which is designed to capture spectral images of the object region; and a control unit (110), which is designed to carry out the following steps: - capturing an image stream of reference images (106) by means of the first capturing strand; - capturing spectral images by means of the second capturing strand (108) at the same time as capturing the image stream of reference images (106); and - generating a representation (112) of the image stream, wherein a sequence image (114) of the object region is superimposed on the representation (112), wherein the sequence image (114) is formed on the basis of the spectral images by the spectral images being locally assigned to the representation (112) of the image stream and by spectral representation parts (116) relating to the spectral images being sequentially added to the sequence image (114) such that the sequence image (115) sweeps over the representation (112) of the image stream.

An apparatus for generating a synthesis audio signal using a patching control signal has a first converter, a spectral domain patch generator, a high frequency reconstruction manipulator and a combiner. The first converter is configured for convert- ing a time portion of an audio signal into a spectral represen- tation. The spectral domain patch generator is configured for performing a plurality of different spectral domain patching algorithms, wherein each patching algorithm generates a modified spectral representation having spectral components in an upper frequency band derived from corresponding spec- tral components in a core frequency band of the audio signal. The spectral domain patch generator is furthermore config- ured to select a first spectral domain patching algorithm from the plurality of patching algorithms for a first time portion and a second spectral domain patching algorithm from the plural- ity of patching algorithm for a second different time portion in accordance with the patching control signal to obtain the modified spectral representation.

The application discloses a kind of random vibration methods of long-span bridge under spatially varying non-stationary earthquake, technical points include: the motion equation of long-span bridge under spatially varying non-stationary earthquake is constructed by finite element method;Modal analysis is carried out to bridge superstructure;Recursive expression between state response vector and excitation vector is constructed;Each seismic input process is discretely characterized using spectral representation method, and it is mapped to the response of bridge support structure corresponding degree of freedom;The explicit relationship between the total response of superstructure in physical space and random input vector is constructed;Based on the explicit relationship between the total response and random input vector obtained, the time-frequency response statistics of bridge structure are evaluated.The application uses the above-mentioned random vibration method of long-span bridge under spatially varying non-stationary earthquake, overcomes the defects that the existing calculation method is low in efficiency and difficult to fully consider the spatial variability and complete non-stationarity of seismic motion, and is very convenient for computer programming implementation.

The described techniques relate to an improved method for geothermal imaging. The method may include obtaining raw seismic data from seismic receivers distributed with reference to a geological zone and detecting a tremor event from the raw seismic data. The detecting may include generating processed seismic data from the raw seismic data; generating spectral representations; generating, for pairs of seismic receivers, a covariance matrix including a cross-spectra between pairs of the spectral representations; generating a time-frequency representation of the processed seismic data; fitting synthetic resonance spectra to the time-frequency representation of the processed seismic data on a per-time-instance basis; and identifying a subset of the processed seismic data based on the fitting, where the subset of the processed seismic data is representative of the tremor event. After detecting the tremor event, method may include outputting candidate locations representative of a source of the tremor event.