66 results about "Frequency domain decomposition" patented technology

The invention provides a fault line selection method of a low current grounding system using a time-frequency atom decomposition theory. The method comprises the following steps of: based on the time-frequency atom decomposition theory, performing sparse decomposition on zero-sequence current data in a Gabor over-complete dictionary, and then obtaining matched attenuation sinusoidal quantity atoms through optimizing and solving relevant parameters. By the time-frequency atom decomposition method, the disturbance characteristics such as start/stop moments, amplitudes, frequencies and change rules of fundamental wave and each subharmonic can be exactly obtained, and interference signals can be effectively filtered. Energy entropies of the atoms decomposed by time-frequency atoms are arranged from large to small; except from the zero-sequence transient current fundamental wave atom, atom phase angles (polarity) similar with zero-sequence current frequency of each line are compared; if the atom phase angle (polarity) similar with the zero-sequence transient frequency of the line is opposite to that of other lines, the line is the fault line; and if the atom phase angle (polarity) of each line is the same, the fault is bus fault, and the fault line is determined by the comparison result of each frequency phase angle.

The invention discloses a brain electrical signal independent component extraction method based on convolution blind source separation. The brain electrical signal independent component extraction method based on the convolution blind source separation includes concrete steps: building a brain electrical signal independent component extraction system based on the convolution blind source separation, which comprises an AD (analog to digital) sampling module, a short time Fourier transformation module, a frequency domain instantaneous blind source separation module, a sequence adjustment module and a short time inverse Fourier transformation module; using the AD sampling module to sample brain electrical signals; using the short time Fourier transformation module to transform the brain electrical signals from a time domain to a frequency domain; using the frequency domain instant blind source separation module to separate instantaneous mixing signals in the frequency domain; using the sequence adjustment module to perform sequence adjustment on independent components in a vector on each frequency domain segment; using the short time inverse Fourier transformation module to transform a frequency domain separation result into an independent component on the time domain. The brain electrical signal independent component extraction method based on the convolution blind source separation extracts the independent components of brain electrical signals based on a true convolution mixing model, uses a convolution blind source separation frequency domain algorithm, and is simple to achieve, good in separation effect, and low in calculation complexity.

The present invention relates to an instantaneous frequency extraction method of Doppler signals, and belongs to the signal processing technology field. According to the present invention, for the Doppler signals outputted by various interference velocimeters, firstly the appropriate parameters, such as the wavelet, the threshold type, the decomposition layer number, the vanishing moment, etc., are selected to carry out the wavelet de-noising on the output signals to smooth the edges; after then the amplitude modulation-frequency modulation decomposition is carried out on the de-noised signals in an iteration manner, and the envelopes of the signals are extracted and are normalized; and finally, the instantaneous frequencies of the normalized signals are extracted by a direct orthogonal method, and an estimation curve of the instantaneous frequencies is obtained in a least-square fitting manner. According to the present invention, firstly the signals are de-noised, so that the direct orthogonal method which is more sensitive to the noise originally can be applied to the Doppler signals, and the simulation results show that on the condition of selecting the appropriate parameters, an error of an instantaneous frequency value obtained by the present invention and a theoretical value is less than 0.5%, the calculation time and the sampling point number are positively related, and the calculation velocity is faster.

The invention relates to a detection method of direct-current bias in a transformer. The method comprises following steps of signal acquisition (S52), signal processing (S54) and judging (S56). The signal acquisition step comprises the step of acquiring a vibration signal of a fuel tank of the transformer. The signal processing step comprises steps of forming a time domain waveform of the vibration signal (S542); carrying out frequency domain decomposition on the time domain waveform so as to transform the time domain waveform into a frequency domain waveform (S544); and calculating an amplitude of each odd number frequency and an amplitude of each even number frequency of the frequency domain waveform (S546). The judging step comprises the step of determining that there is direct-currentbias in the transformer if the amplitude of one odd number frequency is larger than half of the amplitude of one odd number frequency. According to the invention, the detection method is low in detection cost, safe, fast in speed and high in accuracy.

The invention provides a frequency domain convolution blind signal separation method based on multi-objective optimization, which is used for solving the problem that convergence to regressive solution easily occurs in the prior art, and can realize the frequency domain convolution blind signal separation with the number of source signals being smaller than that of observation signals. The methodcomprises the following implementation steps: acquiring a target matrix set R; constructing a diagonalizable matrix B (omega k); constructing a non-orthogonal joint diagonalizable multi-objective optimization model; by utilizing the non-orthogonal joint diagonalizable multi-objective optimization model, estimating the disjunct matrix W (omega k) on each frequency point of the target matrix set R;and acquiring the estimated value of time domain source signals. The method is high in reliability and wide in application range, and can be applied to blind separation of the convolution mixed signals including voice signals, communication signals and the like under over-determined conditions.

The invention relates to a classification method of an electroencephalogram signal. The method comprises the following steps: 1, decomposing an electroencephalogram signal into sum of intrinsic mode functions; 2, performing experience amplitude modulation-frequency modulation decomposition to each intrinsic mode function to obtain an experience frequency modulation component; 3, judging whether the obtained experience frequency modulation component contains a riding wave or not; 4, removing the riding wave if the riding wave is contained in the experience frequency modulation component; 5, calculating an experience amplitude modulation component; 6, calculating an orthogonal component of the experience frequency modulation component; 7, calculating an instantaneous phase; 8, calculating an amplitude modulation bandwidth and a frequency modulation bandwidth; and 9, classifying the electroencephalogram signals by taking the amplitude modulation bandwidth and the frequency modulation bandwidth as input of a support vector machine. The method is not restricted by Hilbert transform of signal product, avoids generation of new riding waves, has good local characteristics, and is improvement to defects of a conventional electroencephalogram signal classification method.

The invention provides a video measurement-based natural vibration frequency identification method for a dam. The method is time-saving, labor-saving and cost-saving, and can provide relatively high space resolution. The video measurement-based natural vibration frequency identification method for the dam is characterized by comprising the following steps: acquiring a natural vibration frequency of an unmanned plane through a wireless acceleration sensor mounted on the unmanned plane; controlling the unmanned plane to fly above a dam top, and shooting a dam top surface by using a camera mounted on the unmanned plane to shoot the edge part of the dam top, so as to acquire a dam top vibration video; extracting a stable video including a dam top edge image from the dam top vibration video; processing the stable video based on a phase motion estimation algorithm to acquire dam top edge motion information; processing the dam top edge motion information through a frequency domain decomposition method, extracting a vibration frequency, and culling out the natural vibration frequency of the unmanned plane from the extracted vibration frequency, so as to obtain a natural vibration frequencyof the dam.

The invention discloses an automatic identification method for a terminal phase sequence. The method comprises the following steps that 1, a basic electrical parameter voltage value u and a current value I when a terminal stably works are collected; 2, collected voltage and current signals are subjected to frequency domain decomposition to obtain base wave amplitude frequency characteristics of voltage and current; 3, the phase angle of a time moment corresponding to voltage and current waveforms in a base wave is analyzed, and whether or not the voltage and current are in the same phase sequence. According to the method, the three phase of the current and voltage can be judged just through a phase angle difference, and the method has the advantages of being simple, economical, easy to apply and popularize and the like.

The invention discloses a dynamic power analysis method based on frequency domain interpolation. The method comprises the following steps: a, discretely sampling actual voltage signals and current signals to obtain a to-be-analyzed sine signal sequence X(n); b, performing fast Fourier transform on the to-be-analyzed sine signal sequence X(n), and calculating spectral interpolation coefficients delta<+> and delta<->; c, calculating weighted spectral interpolation coefficient shown in the description according to amplitudes of two spectral lines adjacent to peak spectral lines; d, calculating the frequency, amplitude and phase of the to-be-analyzed sine signal sequence X(n); e, obtaining the amplitude and phase angle of a voltage sequence to be recorded as Au and theta u respectively, and obtaining amplitude, phase angle and dynamic power of a current sequence. The method solves the problem of low interpolation result precision caused by the fact that when the frequency is calculated with a windowed interpolation FFT algorithm, errors caused by asynchronous sampling or non-integer period truncation of data exist and influence on part of the negative frequency is ignored even if the interpolation algorithm is adopted.

The invention provides a short-term load prediction method based on frequency domain decomposition and an artificial intelligence algorithm, to solve the problems that an existing short-term load prediction method is single in prediction method and low in prediction precision. The method comprises the following steps: decomposing a load time sequence of original load data by using a frequency domain decomposition algorithm to obtain a daily period component, a weekly period component, a low-frequency component and a high-frequency component; predicting a day period and a weekly period by adopting a neural network algorithm; predicting the low-frequency component by adopting a random forest algorithm; and carrying out secondary decomposition on the high-frequency component, and predicting the decomposed low-frequency part by adopting a neural network algorithm. According to the short-term load prediction model based on frequency domain decomposition provided by the invention, compared with an Elman neural network and a random forest prediction result, the prediction result has higher prediction precision.

In an example embodiment, a dynamic feature extraction tool receives a data set from a SHM system that includes a plurality of sensors affixed to a structure (e.g., a bridge, dam, building, etc.), the data set including at least one of ambient vibration data or earthquake vibration data. A solution method is selected from among, for example, time domain analysis, frequency domain decomposition or eigensystem realization analysis. The dynamic feature extraction tool guides a user to select at least one parameter value used in the selected solution method from a subset of determined-effective parameter values computed by the software tool. The dynamic feature extraction tool then automatically performs the selected solution method on the data set using the selected at least one parameter value to determine dynamic features (e.g., frequencies or modal shapes), and displays a graphical representation of the dynamic features in a UI.

The invention relates to a high-permeability photovoltaic power distribution network partition voltage regulation method. The method comprises: carrying out frequency domain decomposition on load datato obtain a daily period, a week period, a low-frequency component and a high-frequency component, respectively calculating each component and then carrying out sequence reconstruction; carrying outabnormal point detection on the photovoltaic data, carrying out similar day clustering selection based on irradiance characteristics, and predicting photovoltaic short-term output power through an LSTM neural network model; dividing the distributed power distribution network into a plurality of sub-communities, and selecting key nodes as installation points of controllable photovoltaics; and selecting controllable PV nodes to be used as key nodes, and adjusting the node voltage after grid connection by adjusting the reactive compensation and active attenuation power values of the controllablePV nodes. According to the method, the prediction precision is improved, refined voltage regulation can be carried out on the distribution network nodes, the problem of voltage out-of-limit after photovoltaic access is solved, and the photovoltaic absorption capacity of a power grid is improved.

The invention relates to a time series data feature extraction method and device, equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the steps: acquiring a time domain signal corresponding to time series data of a user, wherein the time series data represents economic behavior data of the user; carrying out N-stage frequency domain decomposition on the time domain signal by using N stages of filter banks to obtain a plurality of time sequence sub-band signals of different frequency bands, wherein N is a positive integer greater than or equal to 1; and performing feature extraction on each time sequence sub-band signal to obtain target features of the time sequence data. According to the method, the performance of the time series data in different frequency bands can be obtained, and more features in the time series data can be extracted.

The invention discloses a face feature identification method and system based on frequency domain division. The face feature identification method comprises carrying out FFT transformation on a training image to obtain frequency distribution; dividing the frequency distribution to obtain a plurality of frequency components; carrying out IFFT transformation on each frequency component to obtain image components corresponding to different frequency components; combining with a tag of the training image to obtain processed training data; training a convolutional neural network by means of the processed training data to obtain a network parameter; processing an input image to be identified through a model to obtain a feature of the image to be identified; carrying out sample comparison of the image to be identified by calculating the Euclidean distance of the feature to obtain a face feature, and completing face identification. The face feature identification method based on frequency domain division is fast in face identification speed, high in accuracy, strong in robustness and good in noise resisting ability.

This invention relates to roadway identification method by traffic flow test radars, which emits the FM continuous waves generated by a transceiver component, which reflect back when meeting with targets then to be mixed and input to the IF circuit to be amplified and filtered to be input into a signal processor, which carries out roadway identification process to the sampled signal and utilizes a set of filter group same to the roadway number to separate information on different roadways and analyzes the spectrum of each set of data, the transceiver component is mounted on a fixed position with relative fixed background of covered region so as to finish the elimination algorithm of clutters, which engages in frequency domain elimination to spectrums when cars passing through and the spectrums when no cars passing through and records times of cars passing through different ways to get the traffic flow, the signal processor transmits the statistic flows and car velocities to the system controller for network communication.

The invention discloses an image steganalysis method and system based on frequency domain analysis, and the method comprises the steps: carrying out the frequency domain decomposition of a filter group employed by a spatial domain steganalysis model, obtaining a comprehensive frequency domain obtained through the addition of coefficients of all frequency bands of all filters in the filter group, and generating a spectrogram according to a combined frequency domain; obtaining the energy values corresponding to different frequency bands according to the spectrogram, and generating a distortion function at different frequency bands with different weight values according to the difference of energy values; embedding secret information into an original image according to the distortion function, and obtaining hidden carrier information. The method can protect the secret information embedded into the image from being detected by a steganalysis detection algorithm, improves the steganalysis safety of the image, and facilitates the improvement of the communication safety.

The invention discloses a watermark embedding and extracting method and device based on a computer vision technology, computer equipment and a storage medium. The method comprises the following steps: identifying a first feature point of a target image; determining a watermark embedding area of the target image based on the distribution information of the first feature points on the target image; performing frequency domain decomposition processing on the watermark embedding area to obtain at least two sub-band images, and selecting a target sub-band image according to high-frequency information and low-frequency information of the watermark embedding area contained in each sub-band image; embedding the watermark into the target sub-band image to obtain an embedded target sub-band image; replacing the target sub-band image of the watermark embedding area with the embedded target sub-band image; according to the method, the subband image of the watermark embedding area is subjected to inverse processing of frequency domain decomposition processing to obtain the target image with the embedded watermark, so that the watermark embedding area is selected based on the feature points of the target image, the self-adaptability of the watermark embedding area is improved, and the robustness and the safety of the embedded watermark are ensured.

The invention provides a parallel method for solving temperature field by improved SSORPCG of finite element domain decomposition, which comprises the following steps: (1) establishing a solution model, carrying out finite element discretization on the model, and obtaining structural information of the model; (2) dividing the model into a certain number by using domain decomposition method under serial computation, and distributing the unit information and message passing index of each partition; (3) the element heat conduction matrix, the contribution matrix of exothermic boundary to the heatconduction matrix and the element temperature load array being formed independently in each processor, thus the governing equation of finite element being formed; (4) improved SSORPCG solver being called to solve the temperature of each unknown node, and the temperature field of each partition being obtained; (5) the calculation results of the whole model being obtained. Advantages: the inventioncan effectively improve the ability of solving large-scale sparse coefficient matrix equation group of the temperature field, and can improve the quality and efficiency of the finite element analysisand parallel calculation of the temperature field.

The invention discloses a typical power consumption mode extraction method suitable for massive category imbalance load data. The method comprises the steps of (S1) processing load data by adopting aBorderline-SMOTE training sample category imbalance processing method; (S2) decomposing the load data by using MODWT to obtain wavelet coefficients and scale coefficients, and forming a frequency domain characteristic matrix by using the wavelet coefficients and the scale coefficients; (S3) carrying out modeling processing on a frequency domain characteristic matrix obtained after decomposition based on a load classification model of a deep LSTM network; and (S4) carrying out structure parallelization on the load classification model based on Spark. Through the above scheme, the above scheme is adopted, according to the invention, the classification precision of the morphological similarity curve is improved by means of frequency domain decomposition, sample oversampling processing, distributed calculation and the like; the classification precision of the load data with the class imbalance problem is improved, the calculation efficiency of typical power utilization mode extraction of massive load data is improved, and the method has very high practical value and popularization value.

The invention discloses a sound emitting source positioning method, which comprises the following steps of picking lead fracture damage sound emission signals by a sound emission sensor; selecting direct reaching waves in the sound emission signals; performing time domain operation on the direct reaching waves in the sound emission signals; performing frequency domain operation on the direct reaching waves after the execution of the time domain operation; performing displacement domain conversion operation on the direct reaching wave subjected to the execution of frequency domain operation; mapping the signals from the frequency domain to the wave number domain of the signal; mapping the signals of the direct reaching waves subjected to the execution of displacement domain conversion operation to the displacement space; extracting a peak-peak amplitude value and a wave pack length of the signals mapped to the displacement space so as to obtain the change rule of the peak-peak amplitudevalue and the wave pack length along with the phase drift; according to the change rule, judging whether the length of the wave pack is minimum or not when the peak-peak amplitude value reaches the maximum value; determining the position of the sound emission source according to the condition that the transverse moving value on the final time axis dimension is the length of the sound emission source to the sound emission sensor.

The invention discloses a fracture detection method and system based on time-frequency decomposition, and relates to the field of oil and gas physical geography. The method comprises the following steps: obtaining post-stack seismic data of a target layer segment; performing time-frequency decomposition processing on the post-stack seismic data to obtain an amplitude spectrum that changes with thedominant frequency; and performing fracture detection processing on the target layer segment according to the amplitude spectrum that changes with the dominant frequency to obtain a fracture distribution condition of the target layer segment. According to the fracture detection method and system disclosed by the invention, an adaptive spectrum analysis method is imported to serve as a prior processing process based on the calculation of the geometric property of the fracture detection to obtain the amplitude spectrum that changes with the dominant frequency, and the fracture detection processing is performed on the target layer segment according to the amplitude spectrum that changes with the dominant frequency to achieve the fast and accurate detection and identification of formation fracture that changes with the depth.

The invention discloses a frequency domain decomposition based single image defogging acceleration method, which specifically comprises the steps of first, acquiring a foggy image; second, performing k-layer wavelet decomposition on the foggy image, and acquiring a low-frequency component I1 and k high-frequency components I<h><1>, I<h><2>,...,I<h><k>; third, performing defogging processing on the low-frequency component I1 by using an image defogging method so as to acquire a defogged low-frequency component; fourth, processing the k high-frequency components respectively by using a size adjustment model so as to acquire k new high-frequency components; and fifth, performing wavelet reconstruction on the acquired defogged low-frequency component and the k new high-frequency components so as to acquire a defogged image. Compared with an existing image defogging method, the method disclosed by the invention is short in calculation time, high in calculation efficiency and suitable for being applied to real-time image defogging processing.

The invention discloses a self-adaptive anisotropic divided frequency partition filtering method based on energy frequency band distribution. The self-adaptive anisotropic divided frequency partitionfiltering method comprises the following steps: (1) inputting two-dimensional post-stack seismic data u; (2) performing frequency domain decomposition on the two-dimensional post-stack seismic data byutilizing VMD to obtain IMF profiles uk (k is equal to 1, 2 to n) within different frequency ranges, wherein n is an integer; (3) respectively performing multiple iterative processing on each of thedecomposed IMF profiles uk (k is equal to 1, 2 to n) through a self-adaptive threshold anisotropic filtering algorithm and obtaining final refactoring results after each iterative processing and respectively calculating signal-to-noise ratio SNR and similarity SSIM; (5) selecting the optimal final result corresponding to the SNR and the SSIM for output. The self-adaptive anisotropic divided frequency partition filtering method disclosed by the invention has the benefits that de-noising is carried out in combination with the characteristics of a signal in a frequency domain and a time domain, local features and main structure information of seismic data textures are better protected while the noise is filtered, moreover, the seismic data quality is improved.

The invention discloses a transformation ratio quick calculation method for a special transformer in unbalanced state. The method comprises the following steps of A, outputting a three-phase voltage signal by means of a built-in three-phase power supply and applying the voltage signal to the primary side of the transformer; B, calculating a transmission function between two random phase voltage signals at the primary side; C, performing frequency domain decomposition on total three transmission functions; D, selecting transmission function components at the same characteristic frequency segment for performing characteristic vector extraction, and combining the characteristic vectors which belong to different transmission functions; E, forming a characteristic matrix by means of the characteristic vectors, and calculating a characteristic value of the characteristic matrix; F, acquiring a voltage signal at the secondary side of the transformer, and calculating the characteristic value of a voltage signal characteristic matrix at the secondary side of the transformer according to the steps B-E; and G, correcting the voltage signals at the secondary side, and performing calculation for obtaining the transformation ratio of the transformer. The transformation ratio quick calculation method can settle defects in prior art and overcomes a detection error caused by unbalance of a three-phase power supply.