87 results about "Time frequency decomposition" patented technology

A method of determining sleep stages from signals of electrical activity recorded of a chest of a sleeping subject, the signals being measured over a plurality of epochs. The method comprising: (a) extracting a series of cardiac R-R intervals from the signals and obtaining a time-frequency decomposition from the series of cardiac R-R intervals; (b) using the time-frequency decomposition to determine at least one Slow-Wave-Sleep (SWS) period and at least one Non-SWS (NSWS) period; (c) from the at least one NSWS period, determining at least one sleep-onset (SO) period and a plurality of non-sleep periods; (d) extracting a plurality of electromyogram (EMG) parameters from a portion of the signals, the portion corresponds to a NSWS period other than the at least one SO period and other than the plurality of non-sleep period; (e) using the plurality of EMG parameters to determine at least one REM period thereby also to obtain also at least one light-sleep (LS) period defined as a NSWS period other than the SO periods, other than the non-sleep periods and other than the REM periods; thereby determining the sleep stages of the sleeping subject.

The invention discloses a time-frequency slice analysis-based rolling bearing fault diagnosis method. The method comprises the following steps: performing time-frequency decomposition on a vibration signal of a rolling bearing by using frequency slice wavelet transform to obtain a time-frequency energy distribution matrix of the vibration signal; then defining a kurtosis index, i.e., frequency amplitude kurtosis, on the basis of the time-frequency energy distribution matrix; solving the amplitude kurtosis corresponding to each frequency in sequence to form a frequency amplitude kurtosis spectrum of the vibration signal in a time-frequency plane; selecting a plurality of resonant frequency bands on the amplitude kurtosis spectrum; separating resonant frequency band signal components by using frequency slice inverse wavelet transform through a reconstruction method; extracting an envelope component of each resonant frequency band and performing rule processing by using a demodulation analysis method respectively; solving the comprehensive envelope spectrum of the envelope components on the basis; finally recognizing the damage characteristic frequency of the rolling bearing by using the comprehensive envelope spectrum to realize fault diagnosis of the rolling bearing.

A reservoir imaging technique based on full-frequency-band seismic information mining is an oil seismic exploration data processing and interpreting technique, and utilizes the time-frequency decomposition method-the third type generalized S conversion which can accurately describe the local hierarchical structure of seismic signals. Firstly, an original three-dimensional seismic data volume is mapped into a four-dimensional full-frequency-band time-frequency energy data volume, a time-frequency amplitude data volume and a time-frequency phase data volume which all contain time, space and frequency domain; a vertical seismic profile, a time slice, a horizon slice and a stratum slice are extracted from two data volumes by utilizing the geological horizon information and the drilling and logging information; and meanwhile, on the basis of the above data volumes, a full-frequency-band energy difference slice and a reservoir thickness relative to time detection slice which are based on full-frequency-band information are further generated. The technique provided by the invention not only utilizes the information in a passband during conventional seismic data processing but also explores low-frequency and high-frequency information outside the passband, and is used for directly indicating the oil-gas reservoirs and analyzing imperceptible changes of the thickness, the space distribution and the internal structure of reservoirs, thereby improving not only the utilization ratio of the information in the seismic exploration data but also the reliability of seismic data interpretation.

The invention relates to the technical field of petroleum exploration, in particular to a time-frequency decomposition earthquake-fluid recognition method which includes establishing a time-frequency atom dictionary D according to Morlet wavelet function m(t)=exp[-betaXf2(t-tau)2]exp[i(2pif(t-tau)+phi)], and acquiring an initial matching atom of the Morlet wavelet function through calculation with a seismic trace and complex seismic trace method; performing matching decomposition on the seismic trace, performing iterative optimization with constraints of the time-frequency atom dictionary D in the neighborhood of the initial matching atom to acquire an optimal matching atom, stopping matching decomposition when preset conditions are achieved, and representing the initial seismic trace as a series of linear combinations of Morlet wavelet atoms; transforming the optimal matching Morlet atom into the time-frequency domain so as to acquire a time-frequency spectrum distribution of the initial seismic trace; extracting directly properties of earthquake fluid activity on a target stratum section on the time-frequency spectrum of earthquake materials; and predicting distribution range and space distribution of gas deposit according to the properties of the fluid activity. By the method, the distribution range and the space distribution of the gas deposit can be accurately predicted, so that a technical support is provided for favorable target optimization of natural gas exploration.

A method of determining sleep stages from signals of electrical activity recorded of a chest of a sleeping subject, the signals being measured over a plurality of epochs. The method comprising: (a) extracting a series of cardiac R-R intervals from the signals and obtaining a time-frequency decomposition from the series of cardiac R-R intervals; (b) using the time-frequency decomposition to determine at least one Slow-Wave-Sleep (SWS) period and at least one Non-SWS (NSWS) period; (c) from the at least one NSWS period, determining at least one sleep-onset (SO) period and a plurality of non-sleep periods; (d) extracting a plurality of electromyogram (EMG) parameters from a portion of the signals, the portion corresponds to a NSWS period other than the at least one SO period and other than the plurality of non-sleep period; (e) using the plurality of EMG parameters to determine at least one REM period thereby also to obtain also at least one light-sleep (LS) period defined as a NSWS period other than the SO periods, other than the non-sleep periods and other than the REM periods; thereby determining the sleep stages of the sleeping subject.

The invention relates to a distribution network fault classification method based on a convolution depth confidence network. The method comprises the steps of firstly collecting the three-phase voltage, zero-sequence voltage and three-phase current of a low-voltage bus of a main transformer and a low-voltage side of the main transformer, and respectively interceptting the signal waveform data of one cycle wave before and after each fault condition as training samples; secondly, carrying out the time-frequency decomposition on the training sample data of step S1 by using the discrete wavelet packet transform, and obtaining the time-frequency matrix, then constructing the pixel matrix of the time-frequency spectrum map, and constructing the time-frequency spectrum map as the input of the subsequent CDBN model; then constructing the CDBN model to train two convolution-constrained Boltzmann machines in unsupervised learning mode, and adding the softmax classifier after the second CRBM to train the network model to effectively extract and automatically classify the fault features, and finally, using the trained model to realize the fault classification of distribution network. The method of the invention can realize accurate fault location.

The invention discloses a seismic signal time-frequency decomposition method based on short-time fractional Fourier transform. A Gaussian window function is selected; high frequency resolution is obtained by adjusting the window width in a self-adapting way; through the characteristic of fractional kernel function angle rotation, the best energy accumulation of the seismic signal is obtained at the optimal rotation angle and the cross term is eliminated; the higher frequency resolution and more accurate frequency location are more conductive to recognizing specific geologic structure, so that the method solves the defects of the existing seismic signal time-frequency decomposition method. The method has the positive effects of overcoming the defects of traditional algorithm such as short-time Fourier and Wigner-Ville distribution and the like and having great significance on seismic signal spectral analysis.

The invention relates to the technical field of power load monitoring, in particular to a non-invasive load identification model training method, a load monitoring method and a device. The monitoringmethod comprises the following steps: acquiring the original load monitoring data at the entrance of the user's electric power through the Internet of Things equipment, and acquiring the usage situation of the electric appliance; processing The original load monitoring data by noise reduction and time-frequency decomposition. Selecting part load monitoring data, obtaining the corresponding time data and frequency data as training samples, taking the use of electrical appliances as the identification result, constructing the load identification model by training convolutional neural network. The non-invasive load identification model training method of the invention trains the convolution neural network according to the load monitoring data after the time-frequency decomposition, The non-invasive load monitoring method and system of the present invention input load monitoring data based on the load identification model to identify the usage of consumer electrical appliances, thereby improving the accuracy and efficiency of load identification.

The invention discloses a method for diagnosing mechanical states of a distribution switch on the basis of empirical mode decomposition (EMD) sample entropy and fuzzy c-means clustering (FCM). The method comprises the following steps that a piezoelectric accelerometer is mounted near to a main shaft of a switch operating mechanism, and the vibration signals of the distribution switch of different mechanical states at the off and on moments are collected through a data acquisition card and a waveform display storage platform; EMD time frequency decomposition is carried out on the collected vibration signals to obtain the limited number of intrinsic mode components of different frequency and unequal bandwidths and residual error components; the sample entropies of intrinsic mode functions (IMFs) of all orders of the vibration signals of different mechanical states are calculated, and a sample entropy array is formed as the characteristic quantity of the mechanical states of the distribution switch; the sample entropy array is used as the input of the FCM, and the mechanical states of the distribution switch are diagnosed through a fuzzy clustering method.

The invention discloses a high-precision synchronous extraction generalized S transform time frequency analysis method. The method comprises the steps of firstly performing generalized S transform ona signal to obtain a generalized S transform value, thereby obtaining a time frequency spectrum of the generalized S transform; based on the generalized S transform value, calculating an instantaneousfrequency, obtaining a synchronous extraction operator (SEO) according to the instantaneous frequency, extracting the time frequency spectrum of the generalized S transform by utilizing the SEO, onlyreserving energy near a time frequency ridge line in a time frequency plane, removing all other divergent energy, and obtaining a synchronous extraction generalized S transform value; and finally converting the synchronous extraction generalized S transform value into a time frequency spectrum of synchronous extraction generalized S transform. The size of a window function can be flexibly adjusted according to the actual need; the method has higher time frequency decomposition precision and relatively high flexibility; and the time and frequency resolution of the signal is greatly increased.

The invention discloses a seismic attenuation qualitative estimation method based on self-adaptive optimal kernel time frequency distribution. The method comprises the following steps of 1) collecting original seismic data; 2) picking a position of a seismic data target layer; 3) determining a high frequency fH and a low frequency fL; 4) calculating self-adaptive optimal kernel time frequency distribution for each signal in the superimposed seismic data body to obtain a high-frequency component and a low-frequency component; 5) determining a correction factor; and 6) estimating the seismic attenuation. According to the method, the self-adaptive optimal kernel time frequency distribution is firstly used for carrying out the time frequency decomposition of the seismic data to obtain the high-frequency component and the low-frequency component of the seismic data, and then the seismic attenuation is qualitatively estimated by utilizing the difference of the low-frequency component and the high-frequency component. Compared with the traditional seismic attenuation qualitative estimation method based on the time frequency distribution, the method has high time-frequency resolution ratio, and the seismic attenuation can be more precisely reflected; and in addition, the technical scheme is easy to realize, and the operability is strong.

The invention discloses a test method of periodic transient characteristics in signals, uses a sensing device for inputting and implementing analog-to-digital conversion, obtains a signal x(t), and detects whether the transient characteristics with a period T1, T2 or Tm exist in the signal x(t), where m is a positive integer. The test method is characterized in that the test method comprises the following steps: the time frequency decomposition of the signal x(t) is implemented, and the time frequency representation TFRX(t, f) of the signal characteristics is obtained; m polar coordinate mappings are respectively built according to the periods T1, T2 or Tm; each mapping is represented on a polar coordinate figure; when enhanced characteristic representation appears in the polar coordinate figure that is corresponding to the period Tn (n lies in a closed interval between 1 and m), whether the transient characteristics with the period Tn exist in the detected signal is decided. The test method of the periodic transient characteristics in signals conveniently realizes period deciding, improves the efficiency and the accuracy of period deciding, can realize automatic detection of the periodic transient characteristics, and is particularly applicable to automatic failure identification of mechanical equipment.

The invention discloses a time frequency decomposition and reconstruction method of generalized S transform signals for synchronous extrusion. Firstly, the four parameter generalized S transform is conducted on a signal, the variation trend of a basic wavelet function is adjusted by the adjustment of the basic wavelet amplitude, energy attenuation rate, energy delay time and basic wavelet video rate; secondly, the modulus of the result of the four parameter generalized S transform is obtained, and the energy of each time-frequency point is obtained, thus the time-frequency spectrum is obtained; thirdly, the instantaneous frequency is obtained by the four parameter generalized S transform of the signal; fourthly, the frequency set after the generalized S transform is set as the central frequency set, each time frequency point corresponding to the instantaneous frequency close to the interval of the central frequency is extruded to the central frequency point, and the generalized S transform of synchronous extrusion is obtained; finally, the inverse transformation formula of generalized S transform for synchronous extrusion is deduced. The generalized S transform of synchronous extrusion has the advantages of synchronous extrusion transform and generalized S transform, and is a time-frequency decomposition and reconstruction method for high precision signals.

The invention relates to an instantaneous frequency stability based SWT logistics equipment vibration signal noise reduction method in allusion to problems that the material transmission accuracy is reduced and the operation risk is increased because of failure problems such as deviation and abrasion of the traditional transmission equipment due to long-term frequency use. The method comprises thesteps of (1) collecting vibration signals by using a vibration acceleration sensor; (2) performing segmentation processing at intervals on the obtained vibration signals according to a time sequenceto obtain a sampling point set; (3) performing time-frequency decomposition on the vibration segment signals obtained in the step (2) by using synchronous extrusion wavelet transform to obtain a synchronous extrusion wavelet variable and time-frequency diagram information; (4) performing selective extraction and reduction on the variable obtained in the step (3); and (5) calculating an instantaneous frequency curve of an IMT component obtained in the step (4) by using Hilbert transform. The scheme of the invention is applied to the instantaneous frequency stability based SWT logistics equipment vibration signal noise reduction method.

The invention discloses an improved electric vehicle charging non-linear load harmonic wave electric energy metering method. On the basis of the Fourier transform algorithm, the method is performed as follows: using a time-scale analysis method to construct a data analysis algorithm which is consistent with the charging characteristic distribution of an electric vehicle; checking the localized characteristics of a non-stable signal; adjusting the widths of the time window and the frequency window to realize the variable time-frequency resolution; optimizing the electric vehicle non-linear load harmonic wave electric energy metering method; identifying and separating the steady-state harmonic waves and non-steady-state harmonic waves; conducting time-frequency decomposition, stratification and reconstruction to the signal; and achieving the purpose of accurate metering in the entire process. The method of the invention can separate the steady-state and non-steady-state harmonic waves generated by the nonlinear load during the charging operation of the electric vehicle, and performs the multi-scale refinement analysis of the electric power harmonic signal by the telescopic and translational operation of the time scale, which solves the insufficiency problem in Fourier transform in electric vehicle charging non-linear load harmonic wave electric energy metering, therefore, raising the accuracy of electric vehicle charging energy metering.

The invention provides an ISAR (Inverse Synthetic Aperture Radar) imaging method for a complex moving target. Through polynomial phase optimization estimation on each dominant scatterer distance unitecho signal after translational compensation and polynomial phase signal time frequency decomposition, each signal component obtained by decomposition is a single component only corresponding to one frequency point at any time, the defect that cross terms exist in a non-signal component corresponding to multiple frequency points at one time in the traditional time frequency transform is overcome,each dominant scatterer distance unit echo signal has no any cross term interference and building of time frequency distribution with good time frequency joint resolution is realized finally, and range-instantaneous Doppler imaging is thus obtained. The principle is simple, the operation is convenient, bad influences of cross term interference in the classical time frequency analysis method and losses of the time frequency joint resolution are overcome effectively, the quality and the benefits of nonstationary polynomial phase signal time frequency analysis are effectively enhanced, and a target image with good quality and good resolution is obtained.

The invention discloses a data quality analysis-based signal to noise ratio controllable earthquake frequency-expansion processing method. A generalized S transformation time frequency analysis method is adopted for subjecting earthquake data to time frequency decomposition; amplitude preservation of each time frequency component is analyzed and evaluated via well-to-seismic integration, correction factors are calculated via amplitude preservation difference of different time frequency components, the correction factors of each time frequency component are smoothed in time and space scopes, the smoothed correction factors of each time frequency component are used for correcting a corresponding time frequency component, frequency energy can be compensated, signal to noise ratio constraint conditions are specified, and earthquake data is reconstructed under the specified signal to noise ratio constraint conditions. Via the data quality analysis-based signal to noise ratio controllable earthquake frequency-expansion processing method, amplitude preservation of an original earthquake cannot be lowered via frequency expansion operation, and an earthquake frequency band can be widened to a maximum degree under the specified signal to noise ratio conditions.

The invention provides a time frequency decomposition method for signals including intersections in instantaneous frequencies. The method comprises the following steps of: modulating an original signal by using a linear frequency modulation operator to remove the intersection of the instantaneous frequency momentarily; performing empirical mode decomposition on the signal to obtain an intrinsic mode function; modulating the intrinsic mode function by using the conjugate of the linear frequency modulation operator; performing Hilbert transform to obtain a clear time-frequency distribution diagram. The invention mainly provides a method for improving the signal time frequency decomposition capacity by means of linear frequency modulation wavelet transform specific to signals including intersections in the instantaneous frequencies.

The invention relates to a sound based coal-rock interface recognition device and a recognition method. The device consists of an anti-explosion shell, a core processor arranged in the anti-explosion shell, a data memory, an audio signal receiving module, a signal transmission module and an intrinsic safety type battery. The anti-explosion shell is externally provided with a microphone, the sound signal collected by the microphone is transmitted to the core processor by the audio signal receiving module, the core processor is connected to the data memory, the core processor carries out analytical processing on the audio signal, then the signal transmission module transmits the signal to a coal mining machine controller, and the coal mining machine controller is connected to a coal mining machine hydraulic system execution unit. The method utilizes the time-frequency decomposition characteristics, positioning and masking characteristics of human ear physiology to conduct further processing so as to distinguish the current cutting state of the coal mining machine, thereby quickly recognizing the coal-rock interface. The device and the method provided by the invention have the advantages of high measurement accuracy, cause small device loss, suitability for all geological conditions, and wide application range.

The invention discloses a road sound recognition method and system, computer equipment and a readable storage. The road sound recognition method comprises the steps that data samples and sample categories of road sounds areacquired; time-frequency decomposition, frequency conversion, logarithmic operation and derivation areperformed on the data samples in sequence to obtain logarithmic Mel characteristics of the data samples; the logarithmic Mel characteristics areinput into aconvolution cyclic network model for training according to the sample category until the convolution cyclic network model meets a preset training end condition; and the convolution cyclic network model after training isutilized to identify and classify the voice data to be processed. According to the road sound recognition method and system,the computer equipment and the readable storage, the logarithmic Mel characteristics of the road sound data samples are taken as inputs of the convolution cyclic network model,and the model which can be used for identifying multiple sounds in a complex traffic scene is trained, so that the accuracy of road traffic event detection is improved.

The invention relates to a wavelet decomposition-based earthquake spectrum decomposition method, and includes the steps of: 1) inputting seismic data, and performing Hilbert conversion and spectral analysis on the seismic data; 2) presetting a Ricker wavelet function; 3) building a Ricker wavelet set according to a spectrum of the seismic data obtained by the Step 1) and the Ricker wavelet function in the Step 2); and 4) using the Ricker wavelet set to perform time-frequency decomposition on a seismic trace. The wavelet decomposition-based earthquake spectrum decomposition method has the beneficial effects that time-frequency resolution of earthquake spectrum decomposition can be greatly improved, and high-precision time-frequency domain seismic data can be obtained, thereby laying a solid foundation for improvement of seismic reservoir prediction; and the time-frequency resolution is improved by more than one time, playing an important role in using seismic frequency division information to perform reservoir prediction, and providing an important guarantee for oil companies and service companies to improve the success rate of exploration and development drilling.

The invention discloses an ideal seismic spectrum decomposition method based on variable phase ricker wavelet matching tracking. The method comprises the following steps of firstly, constructing a variable phase ricker wavelet library; then, according to a matching tracking method, calculating a matching wavelet and a corresponding amplitude coefficient, and multiplying the amplitude coefficient by a wavelet atom and taking as a sub-signal obtained through decomposition; and then, carrying out Fourier transform on the sub-signal corresponding to a single reflection interface, calculating the amplitude spectrum and the rotational phase spectrum of an interface reflection seismic wave, and placing a Fourier transform spectrum at corresponding delay time; and finally, acquiring an ideal timefrequency spectrum decomposition result. The time-frequency decomposition result comprises the two parts of an amplitude spectrum and a rotational phase spectrum, and the rotational phase spectrum canprovide richer geological and reservoir information. The amplitude spectrum decomposition result accurately reflects the delay time of a seismic wavelet arriving at an interface and energy distribution along with a frequency change, is not limited by a uncertainty principle limit, has a super high time resolution and a frequency resolution and is an ideal time-frequency spectrum for a seismic signal.

The invention provides a polynomial phase signal adaptive time frequency transformation method based on genetic optimization. Time frequency decomposition of polynomial phase signals can be finished,every signal component obtained by decomposition only corresponds to a single component of a frequency point at any time, then values are obtained by the various signal components and instantaneous frequency at any time, signal frequency distribution corresponding to the corresponding time is directly calculated and generated by only retaining a main lobe responded Sinc function, the shortcoming that cross terms exist due to the fact that one time corresponds to non-single components of a plurality of frequency points in the traditional time frequency transformation is overcome, and thus, timefrequency distribution which does not have any cross term interference and has excellent time frequency combined resolution is output finally. The polynomial phase signal adaptive time frequency transformation method based on genetic optimization is simple in principle and convenient to operate; adverse impact of cross term interference of a classic time frequency analysis method and loss of timefrequency combined resolution can be overcome effectively, and the quality and benefit of non-stable polynomial phase signal time frequency analysis can be improved effectively.

The invention discloses a reflective slot wave frequency dispersion similarity-based tomographic imaging method. The method is realized through the following steps: 1, conducting the time-frequency decomposition; 2, figuring out a through slot wave frequency dispersion curve; 3, eliminating through slot waves; 4, imaging. The method makes full use of the frequency dispersion properties of slot waves, and reduces the influence of other noises on the imaging process based on the frequency dispersion property. Meanwhile, the arc phenomenon is reduced, and the tomographic imaging accuracy is greatly improved.

The invention discloses a realization method of frequency domain amplitude spectrum kurtosis diagram. The realization method comprises the following steps: firstly performing frequency slice wavelet transform on a signal to obtain a time frequency decomposition matrix, on this basis, performing signal reconstruction according to each sub frequency band with given scale, and solving an mean amplitude of a reconstructed signal, performing Fourier transform to obtain a square spectrum of the constructed signal, solving the spectrum kurtosis for an amplitude sequence of the square spectrum, and arranging the spectrum kurtosis diagram values according to scale-frequency band so as to obtain a spectrum kurtosis diagram. The size of the spectrum kurtosis is represented by using the color depth inthe spectrum kurtosis diagram, the larger the spectrum kurtosis value is, the deeper the color of the corresponding spectrum plane region is; in the application, the frequency band corresponding to the plane region with the deepest color is selected to reconstruct so as to obtain the time frequency feature signal for further analysis. The realization method disclosed by the invention can be usedfor signal analysis and fault feature extraction of the mechanical equipment and the electric system; and especially for the signal with the frequency-modulation and amplitude-modulation feature, theoptimal modulation frequency band of the frequency-modulation and amplitude-modulation component can be displayed in the diagram.

The invention relates to a two-channel speech enhancement system and method, and belongs to the field of speech processing. The two-channel speech enhancement system comprises a speech input end used for collecting speech information and a speech processor. The speech input end is used for collecting near-end target speech information through two channels, and the collected speech information is subjected to speech enhancement processing and then is transmitted to a communication equipment terminal used for receiving language signals, wherein speech enhancement processing includes the steps of time-frequency decomposition, power spectrum smoothness, feature extraction, attenuation coefficient generation, signal reestablishment and the like. According to the two-channel speech enhancement system and method, on the basis of the principle of computational auditory sense analysis, two paths of telephone transmitters are used as input, in a near-end speaking mode, good restraining and removing effects are achieved for strong background noise, the overall noise reduction effect is greatly superior to that of a current military 'near-speaking pressure difference' telephone transmitter, the signal to noise ratio can be effectively increased during voice communication, good speech quality is guaranteed, and the problems that the noise reduction effect of the military telephone transmitter in the prior art is poor and the speech quality is influenced are solved.

The invention discloses a bearing fault diagnosis method under a variable working condition based on mixed entropy and joint distribution adaptation. An original sample data set of rolling bearing vibration is divided according to whether the load condition is known to establish a source domain data set and a target domain data set; time-frequency decomposition is carried out on different sample signals in the source domain data set and the target domain data set to obtain an intrinsic mode function of each sample signal; a nonlinear entropy parameter of each IMF component of each sample signal is calculated according to a nonlinear metric entropy theory, and a multi-scale mixed entropy feature vector of the sample signal is constructed; and according to transfer learning theory, the multi-scale mixed entropy feature vectors of all sample signals in the source domain data set and the target domain data set are used for constructing a rolling bearing fault diagnosis model under the variable working condition based on joint distribution adaptation, and a final diagnosis result is output. The method can effectively solve the problem in recognizing various fault states of the bearing under the unknown load condition, remarkably improves the fault diagnosis precision, and improves the stable operation level of the bearing.

The invention provides a polynomial phase signal time-frequency transform method based on particle swarm optimization which is capable of performing time-frequency decomposition on polynomial phase signals; each signal component acquired from the decomposition is a single component corresponding to only one frequency point at any moment; the signal components and instantaneous frequency values ofall moments are subjected to direct calculation by reserving only main-lobe-responsive Sinc function so as to generate a signal frequency distribution of each corresponding moment; the defect is overcome that non-single components with one moment corresponding to multiple frequency points in traditional time-frequency transform have cross components; a time-frequency distribution with no cross-component interference and good time-frequency joint resolution is finally output. The polynomial phase signal time-frequency transform method based on particle swarm optimization has the advantages thatthe principle is simple, operating is convenient, the adverse influence of cross-component disturbances from the conventional time-frequency analysis methods and the loss of time-frequency joint resolution can be effectively overcome, and the quality and benefit of nonstationary polynomial phase signal time-frequency analysis can be effectively improved.

The invention provides a fluid identification method and a fluid identification device, wherein the fluid identification method comprises the steps of acquiring a prestack angle gather data volume of a researched area; performing angled stacking on the prestack angle gather data volume seismic data for obtaining an angled-stacked data volume; performing prestack AVO inversion on the angled-stacked data volume for obtaining a longitudinal-wave impedance body and a density body; performing time-frequency decomposition processing on the longitudinal-wave impedance member for obtaining a dividing frequency impedance body; performing inversion operation on the dividing frequency impedance body according to the density body, thereby obtaining a reservoir seismic frequency dispersion attribute on the researched area; and performing fluid identification according to the reservoir seismic frequency dispersion attribute. The fluid identification method and the fluid identification device settle a technical problem of incapability of accurately and effectively performing fluid identification in prior art and realizes a purpose of accurately identifying fluid.

The invention discloses a fractional order synchronous extraction generalized S transform time-frequency decomposition and reconstruction method. The method comprises the following steps: inputting x(t), selecting different rotation angles alpha to carry out fractional Fourier transform spectral analysis on x (t); selecting an optimal rotation angle alpha opt to carry out fractional order generalized S transformation on the signal to obtain a fractional order generalized S transform value, and performing modeling; solving fractional order instantaneous frequency estimation on the basis of thefractional order generalized S transform value, obtaining a synchronous extraction operator according to the instantaneous frequency estimation, extracting a fractional order generalized S transformtime-frequency spectrum by using the synchronous extraction operator, retaining effective energy, and finally reconstructing a signal. According to the invention, the time-frequency characterization of the signal can be popularized to the time fractional order frequency characterization; time-frequency information details can be recognized by selecting the optimal rotation angle in a self-adaptivemode, on the basis, energy of a fractional order synchronous extraction generalized S transformation time-frequency spectrum is more gathered by adjusting window function parameters and conducting extraction operation, and therefore the time-frequency resolution of signals is greatly improved.