Time Frequency Domain Analysis Methods Research Articles

A Time-Frequency Domain Analysis Method for Variable Frequency Hopping Signal.

A variable frequency hopping (VFH) signal is a kind of frequency hopping (FH) signal that varies both in frequency and dwell time. However, in radio surveillance, the existing methods for unidentified signals using VFH cannot be effectively handled. In this paper, we proposed an improved joint analysis method based on time-frequency domain features, which adopts multi-level processing to solve the time-frequency domain feature analysis problem of the VFH signal. First, the received signal is pre-processed by Short-Time Fourier Transform (STFT) and binarization, and a highly discriminative time-frequency image is obtained; then, the fixed frequency signal is removed based on the feature of connected domains, and the conventional frequency hopping (CFH) signal is removed by density-based spatial clustering of applications with noise (DBSCAN); finally, the overlapping region is cropped by the joint energy peak time-domain continuity properties. After the above multi-level joint processing method, the problem of VFH signal processing is effectively solved. The simulation result shows that the Mean Square Error (MSE) between the output results and the time-frequency image of the original VFH signal tends to be close to 0 when the Signal-to-Noise ratio (SNR) is 5 dB.

Gearbox fault diagnosis based on RGT-MFFIN and multi-sensor fusion image generation

Abstract In gearbox fault diagnosis based on vibration and torque state data, traditional one-dimensional time-frequency domain analysis methods often suffer from insufficient feature expression and mining, and require complex noise reduction and filtering preprocessing. To address this issue, this paper proposes a fusion image generation method that integrates the advantages of recurrence plot (RP) and Gramian angular summation field (GASF) to generate recurrence Gramian transformed (RGT) images. This approach integrates both global and local fault information, making the fault characteristics more intuitive and easier to analyze. Given that multi-sensor collaboration can enhance feature representation, feature-level fusion increases the computational burden, and decision-level fusion is prone to losing inter-sensor correlation information, this paper adopts data-level fusion for image sample enhancement. In the diagnostic method, the challenge of traditional convolutional neural networks (CNNs) in extracting diverse geometric linear structures from fused images is addressed by introducing deformable convolutional blocks for initial feature extraction. Additionally, a multi-scale feature fusion interaction network (MFFIN) is constructed. This network incorporates a channel-space interactive attention mechanism on top of multi-scale feature extraction, assigning weights to features according to their importance while facilitating the interaction of feature information. Finally, validation is carried out using public datasets, and the experimental results show that the proposed method demonstrates significant advantages in classification accuracy and robustness under variable operating conditions and noise, thereby proving its effectiveness and practicality.

Feature Decoupling for Multimodal Locomotion and Estimation of Knee and Ankle Angles Implemented by Multi-Model Fusion.

Many challenges exist in the study of using orthotics, exoskeletons or exosuits as tools for rehabilitation and assistance of healthy people in daily activities due to the requirements of portability and safe interaction with the user and the environment. One approach to dealing with these challenges is to design a control system that can be deployed in a portable device to identify the relationships that exist between the gait variables and gait cycle for different locomotion modes. In order to estimate the knee and ankle angles in the sagittal plane for different locomotion modes, a novel multimodal feature-decoupled kinematic estimation system consisting of a multimodal locomotion classifier and an optimal joint angle estimator is proposed in this paper. The multi-source information output from different conventional primary models are fused by assigning the non-fixed weight. To improve the performance of the primary models, a data augmentation module based on the time-frequency domain analysis method is designed. The results show that the inclusion of the data augmentation module and multi-source information fusion modules has improved the classification accuracy to 98.56% and kinematic estimation performance (PCC) to 0.904 (walking), 0.956 (running), 0.899 (stair ascent), 0.851 (stair descent), respectively. The kinematic estimation quality is generally higher for faster speed (running) or proximal joint (knee) compared to other modes and ankle. The limitations and advantages of the proposed approach are discussed. Based on our findings, the multimodal kinematic estimation system has potential in facilitating the deployment for human-in-loop control of lower-limb intelligent assistive devices.

Machine learning-based classifying of risk-takers and risk-aversive individuals using resting-state EEG data: A pilot feasibility study.

Decision-making is vital in interpersonal interactions and a country's economic and political conditions. People, especially managers, have to make decisions in different risky situations. There has been a growing interest in identifying managers' personality traits (i.e., risk-taking or risk-averse) in recent years. Although there are findings of signal decision-making and brain activity, the implementation of an intelligent brain-based technique to predict risk-averse and risk-taking managers is still in doubt. This study proposes an electroencephalogram (EEG)-based intelligent system to distinguish risk-taking managers from risk-averse ones by recording the EEG signals from 30 managers. In particular, wavelet transform, a time-frequency domain analysis method, was used on resting-state EEG data to extract statistical features. Then, a two-step statistical wrapper algorithm was used to select the appropriate features. The support vector machine classifier, a supervised learning method, was used to classify two groups of managers using chosen features. Intersubject predictive performance could classify two groups of managers with 74.42% accuracy, 76.16% sensitivity, 72.32% specificity, and 75% F1-measure, indicating that machine learning (ML) models can distinguish between risk-taking and risk-averse managers using the features extracted from the alpha frequency band in 10s analysis window size. The findings of this study demonstrate the potential of using intelligent (ML-based) systems in distinguish between risk-taking and risk-averse managers using biological signals.

Milling tool wear prediction: optimized long short-term memory model based on attention mechanism

To improve the prediction accuracy of milling tool wear, a prediction method based on Attention-LSTM is proposed. In the training phase, first, the data are pre-processed by truncation, downsampling, and the Hampel filtering method, and then features are extracted by the time domain, frequency domain, and time-frequency domain analysis methods. Second, a deep neural network is designed to describe the complex nonlinear function between features and tool wear. Last, aiming at the insufficient prediction accuracy due to the LSTM lacking feature extraction and enhancement, the Attention mechanism is introduced to optimize the model. The results suggest that this prediction method provides an efficient strategy for milling tool wear prediction.

Auxiliary harmonic excitation generalized method for higher-order analysis of linear structure under multiple non-stationary non-Gaussian excitations

In the classical random vibration analysis, the frequency-domain analysis method, especially the time-frequency domain analysis method based on the frequency domain analysis method, is the most commonly analysis method. However, the time-frequency domain analysis method is challenging when dealing with multiple non-stationary non-Gaussian excitations. In this paper, an accurate and efficient auxiliary harmonic excitation generalized method for higher-order analysis of linear structures under multiple non-stationary non-Gaussian excitations is proposed. Firstly, the expression for the higher-order moment spectrum of response is theoretically deduced by the mode superposition method, in which the time-varying generalized impulse response function and the time-varying generalized frequency response function are introduced. Secondly, according to the relationship between the response of harmonic excitation and the time-varying generalized frequency response function, the expression for the high-order moment spectrum of response is reconstructed in terms of the response of harmonic excitation and the higher-order moment spectrum matrix of the excitations, in which the response of the time-varying auxiliary harmonic excitation can be evaluated efficiently based on the time-domain explicit formulation method. Finally, two examples are investigated to verify the accuracy and efficiency of the proposed method.

A multibranch residual network for fault-diagnosis of bearings

Time-frequency domain analysis methods are used to diagnose faults in bearings by extracting the features of fault signals. Given that a fault signal is also a form of audio signal, we extracted the characteristics of the mel spectrum from the original signal and applied it to a convolution neural network proposed in this paper. Focusing on the residual structure in the residual neural network (ResNet), we solved the gradient disappearance problem and accelerated the training of the model. The importance of each feature channel could be estimated adaptively using the squeeze-and-excitation network (SENet) considering the relationships between the channels. We examined the feature map of each layer using a multibranch residual network (MB-ResNet) to characterize the bearing fault signal. We used the multibranch residual structure to reduce the sense field of each residual and added a parallel local sensing module to train the model to recognize the weight of each input feature to either increase or reduce the influence of local features. Our experimental results show that the MB-ResNet is very good at extracting features, is robust, and capable of generalization.

Ensemble Learning-Based Pulse Signal Recognition: Classification Model Development Study.

BackgroundIn pulse signal analysis and identification, time domain and time frequency domain analysis methods can obtain interpretable structured data and build classification models using traditional machine learning methods. Unstructured data, such as pulse signals, contain rich information about the state of the cardiovascular system, and local features of unstructured data can be extracted and classified using deep learning.ObjectiveThe objective of this paper was to comprehensively use machine learning and deep learning classification methods to fully exploit the information about pulse signals.MethodsStructured data were obtained by using time domain and time frequency domain analysis methods. A classification model was built using a support vector machine (SVM), a deep convolutional neural network (DCNN) kernel was used to extract local features of the unstructured data, and the stacking method was used to fuse the above classification results for decision making.ResultsThe highest average accuracy of 0.7914 was obtained using only a single classifier, while the average accuracy obtained using the ensemble learning approach was 0.8330.ConclusionsEnsemble learning can effectively use information from structured and unstructured data to improve classification accuracy through decision-level fusion. This study provides a new idea and method for pulse signal classification, which is of practical value for pulse diagnosis objectification.

Numerical Study on Pressure Pulsation in a Slanted Axial-Flow Pump Device under Partial Loads

The 30° slanted axial-flow pump device is widely used in agricultural irrigation and urban drainage in plains areas of China. However, during the actual operation process, the 30° slanted axial-flow pump device is prone to vibration, noise, cracks in the blades, and other phenomena that affect the safe and stable operation of the pump device. In order to analyze the flow pressure pulsation characteristics of the 30° slanted axial-flow pump device under different flow conditions, the time–frequency domain analysis method was used to analyze the pressure pulsation of each flow structure of the 30° slanted axial-flow pump device. The results showed that the internal pulsation law of the elbow oblique inlet flow channel is similar. At the 1.2 Qbep condition, the amplitude fluctuation of the pressure pulsation was small, and the main frequency is 4 times the rotating frequency. The monitoring points at the outlet of the elbow oblique inlet flow channel were affected by the impeller rotation, and the pressure pulsation amplitude was larger than that inside the elbow oblique inlet flow channel. The pressure fluctuation of each monitoring point at the inlet surface of the impeller was affected by the number of blades. There were four peaks and four valleys, and the main frequency was 4 times the rotating frequency. The amplitude of pressure fluctuation increased gradually from the hub to the rim. The main frequency of pressure fluctuation at each monitoring point of the impeller outlet surface was 4 times of the rotating frequency, and the low frequency was rich. The amplitude of pressure fluctuation was significantly lower than that of the impeller inlet. With the increase of flow rate, the peak fluctuation of pressure coefficient decreased gradually, and the amplitude of pressure fluctuation tended to be stable. Under 0.8 Qbep and 1.0 Qbep conditions, the large fluctuation of the pressure fluctuation amplitude on the outlet surface of the guide vane was mainly affected by the low-frequency fluctuation. Under the 1.2 Qbep condition, the pressure fluctuation amplitude changed periodically.

Study on the frequency of acoustic emission signal during crystal growth of salicylic acid

AbstractBased on the results of the previous experiment, this article studied the acoustic emission (AE) signals released during the crystallization of salicylic acid to establish the relationship between the AE signal and the particle size. A tremendous amount of acoustic data was analyzed using time–frequency domain analysis methods in order to extract the valuable contents. Based on the diffusion theory, the vibratory model between the AE signal and the crystal particle size was established. This article mainly studies the process of small particles diffusing to the growth point by impact, adding to the lattice, and the crystal releases energy. The impact of the growth unit on particle aggregate is equivalent to a linear elastic vibration system with one end fixed and the other end free. The vibration frequency is 200–355 kHz when the particle size is between 600 and 1,100 µm. The calculated vibration frequency is in good agreement with the measured frequency.

The Correlation Study of Cun, Guan and Chi Position Based on Wrist Pulse Characteristics

To analyze the correlation of pulse characteristics at different acquisition positions, the pulse parameters that can identify different pulse and physiological information are given. In this paper, the relevance and difference of wrist pulse characteristics at Cun, Guan and Chi position were studied by pulse diagnosis instrument and time-frequency, frequency-domain and time-frequency domain analysis methods etc. Firstly, the pulse information of Cun, Guan and Chi position is collected from 206 healthy adults. The 66 pulse characteristics of each sampled location are extracted. In addition, the 66 feature indexes of each position are used to calculate the Spearman correlation coefficients among Cun, Guan, Chi position as well as the both hands in detail. Finally, the relevance and difference of wrist pulse at Cun, Guan and Chi position were studied and analyzed by significance test. The pulse characteristics with high correlation at Cun-Guan-Chi position of left and right hands are mainly power spectrum characteristics and time-frequency characteristics, while the features with high correlation between both hands mostly concentrate in the time domain characteristics. This work has studied the relationship between the position of pulse acquisition from quantitative and qualitative aspects, and laid a foundation for the objective diagnosis of pulse condition and the comprehensive disclosure of pulse information.

Comparison study of three-component polarization analysis methods for seismic advanced detection in the roadway

Multicomponent seismic exploration makes up for the shortcomings of multiple solutions in conventional single-component seismic exploration for lithological identification and geological structure interpretation. Polarization analysis methods play a key role in the processing of multicomponent seismic data. However, when multiple types of waves are very close to each other, it is very difficult to characterize the polarization distribution of different waveforms using conventional time domain or time-frequency domain polarization analysis. In this paper, a time-frequency domain instantaneous polarization analysis method based on a time-frequency method of generalized S-transform is introduced. The instantaneous polarization parameters (polarization axis, polarizability, and planarization vector) can be calculated based on a time-frequency spectrum and instantaneous phase of multicomponent signal. The numerical simulation of three-component seismic advanced detection in the three-dimensional space shows the following results: the diffracted P-wave from a fault breakpoint has the characteristics of narrow wave train and fast speed, which make it easy to be identified. Besides, it is an effective wave for polarization analysis due to its distinct polarization characteristics. Combining the advantages of the time-frequency domain analysis method, this method is more accurate in waveform identification and more reliable in fault identification. Besides, this method provides better results than the conventional time-domain polarization analysis method.

(Invited) Experimental Characterization and Modeling of Heat Generation Rates of Lithium-Ion Batteries with NMC/C Chemistries

The operating temperature of a battery strongly affects overall chemical reactions, ion transport, intercalation and deintercalation process and consequently efficiency, cycle life, and safety of battery systems. Therefore, battery thermal management system (BTMS) should be optimally designed to ensure a highly efficient and reliable operation of the battery system, which requires characterization and analysis of heat generated during operations rather than temperature.Characterization of heat generation rates of a battery requires a highly accurate and dynamic calorimeter that is not commercially available. Therefore, we designed a new multifunctional calorimeter for lithium-ion batteries using thermoelectric assemblies (TEA), which enables the measurement of heat generation rate as a function of C rates, state-of-charge (SOC) and temperatures. In addition, the calorimeter also works as a thermostat that is capable of tracking an alternating reference temperature, so that flexibility and accuracy of experiments are ensured. This capability enables to decouple the temperature fluctuation during charging and discharging operations and the measurement of the entropy coefficient.In this paper, the cells used as an example for the experiments are large format pouch type lithium ion batteries with LMO/NMC-Graphite chemistry. The lumped heat generation rate of the cell is measured at different C rates, state-of-charge (SOC) and temperatures. In addition, analysis of the data has shown that both the heat generation rate and the ratio of the total heat generation to the total energy stored or retrieved of the tested cells increase as the current increases or the temperature decreases.Characterization of the heat source terms includes measurement of reversible and irreversible heat sources that can be determined by calorimetric methods. Particularly, the entropy coefficient (dUOC/dT) in the reversible source is the key parameter that determines the amount of the reversible heat generated during operations. Currently, either potentiometric or calorimetric method is used to measure the value of the entropy coefficient, which drawbacks are long measurement time and inaccuracy. Thus, a new method is proposed based on the calorimeter that works as a thermostat in conjunction with an estimation technique. In order to reduce the measurement time, a time-domain background voltage offset correction and a frequency-domain entropy coefficient determination have been newly developed and applied. Comparison between the developed one and the classical methods has shown a drastic reduction of the measurement time and a similar accuracy to the potentiometric method. The proposed method is also used to investigate effects of temperatures and aging on the entropy coefficient. As a result, the reversible heat source term dependent upon the operating conditions such as SOC, temperature and cycle life can be characterized, which are the important factors for design of battery thermal management systems.Moreover, a calorimetric method for the characterization of the reversible and irreversible heat sources is also developed by applying a sinusoidal current excitation to the cell and analyzing the heat generation response in frequency domain. The determined parameters of both heat sources are compared with those measured by hybridized time frequency domain analysis method and EIS analysis, respectively, which shows a good match. Moreover, the calculated reversible and irreversible heat sources terms are compared with the experimental measurement. Results have shown improved performances in the characterization of irreversible and reversible heat with respect to measurement speed and accuracy.Finally, a thermal model that includes irreversible and reversible heat source terms is developed and then incorporated into a reduced-order electrochemical model (ROM). The electrochemical thermal model is validated against the heat generation rate of the cells measured by the developed calorimeter aforementioned. The model is capable of predicting heat generation rates up to 3C from -30°C to 35°C. Based on the validated thermal model, effects of C rates and temperatures on heat sources can be analyzed. The results have shown that the reversible heat is dominant at low C rates, while the irreversible one is dominant at high C rates because the Joule heating, which is the main irreversible heat source, is proportional to the square of the current. Besides, the reversible heat source is dominant at high temperatures, while the irreversible one is dominant at low C rates because battery internal resistance sources such as contact resistance and ionic resistivity rapidly increase at low temperatures.In addition, the lumped model is used to develop a two-dimensional electrochemical thermal model that is validated with temperature images captured by an Infrared thermal camera.Presented experimental characterization of heat generation and the associated theoretical analysis by electrochemical thermal models can provide basics and fundamentals for an optimal design of future battery thermal management systems.

Fault Diagnosis of Ball Screw in Industrial Robots Using Non-Stationary Motor Current Signals

With the advancement of intelligent manufacturing, different kinds of industrial robots have been applied in modern factories. The liquid crystal display transfer robot (LCDTR) has been widely used in LCD production lines to transport panels. Effective fault diagnosis and prognosis of the industrial robots are of great importance, since unplanned downtime caused by faulty robots significantly reduces the production capacity. Specifically, the ball screw is the critical component in the LCDTR. The failure of the ball screw can cause long downtime. Conventionally, the fault diagnosis of the ball screw is usually based on the vibration signals. However, it is extremely difficult to install the vibration sensors in the industrial robots. Therefore, in order to address this issue in condition monitoring, this paper proposes a data-driven fault diagnosis methodology using the motor current signals of the ball screw. Two time-frequency domain analysis methods are investigated, including short-time Fourier transform (STFT) and wavelet packet decomposition (WPD). The statistical features are extracted, and Fisher score is used to select features. Furthermore, the logistic regression and k-nearest neighbors are applied for the final fault diagnosis. Experiments on a real-world industrial robot dataset are carried out for validation. 100% diagnosis accuracy can be basically achieved by the proposed method, which indicates the non-stationary current signal can be effectively used to identify the health states of the ball screw in the LCDTR.

An experimental study on the characteristics of vibration source in urban rail transit turnouts

Due to the structural characteristics of railway turnouts, as well as their vulnerability to damage and other problems, the environmental vibrations induced by urban rail transit often have a greater effect on the railway turnouts than that on an ordinary line. Therefore, it is necessary to carry out an in-depth study on the characteristics of vibration source on the turnouts. In this paper, a group of representative turnouts is selected for field testing on four typical lines of different operational times in a Chinese city. The profile and rail irregularity of the turnouts are measured to evaluate the service condition of the turnouts and the wheel–rail interaction. The vibration acceleration of the track structure and tunnel wall is measured as the train passes the turnout by setting the measuring points in the two main vibration sources, including the switch and the crossing section. The time–frequency domain analysis method is used to analyze these data. It is found that the source strength of the tunnel wall in the crossing section is much greater than that in the switch section, and that the lateral vibration level is often greater than the vertical vibration level. Unexpectedly, there is a low positive correlation between the vibration level and the train speed, while the healthy state of the turnout rails has a great influence on the vibration level. According to the results of the research, the sources of vibration excitation in the turnouts are complex; therefore, it is necessary to comprehensively evaluate their environmental impact and take corresponding vibration control measures.

Innovative Methods in Planning, Design and Lighting of Music Visual Buildings in Modern Architectural

Based on the analysis of the data of the landscape live performances which have already been performed, this paper combs the development process of the live performances and other information of the performance types so that the public can have a clearer and clearer understanding of the live performances. Through time-frequency domain analysis method, the characteristic elements of waveform music are extracted and classified into sub-music segments in different rough emotional domains according to certain classification methods. And the application weights of the feature values of each music feature are defined in different emotion domains, which makes the input of the system more rational. For the output of the light action control data, this article has a certain interpretation of the DMX512 data, making the interpretation of the output more intuitive. The basic characteristics of music include speed, mode, beat, tone, volume, and the pitch, pitch, and length of notes. Therefore, in the future, the method of realizing music visualization in architectural lighting design and the method of realizing interaction with people will have more possibilities.

Determination of SS321 pitting stage in FeCl3 solution based on electrochemical noise measurement data using artificial neural network

Artificial neural network (ANN) was used successfully to determine the pitting stage of stainless steel 321 (SS321) based on electrochemical noise (ECN) data. The ANN inputs were the parameters derived from the analysis of ECN data. The target values were the pitting stage of SS321 in FeCl3 solution, which were determined using galvanostatic electrochemical impedance spectroscopy (GEIS) tests. The ANN was validated by using a large number of available experimental data. Four approaches have been used for ANN development. The approaches differ in the number and type of inputs. The results showed that the best performance of ANN was achieved when all ECN analysis parameters were used as ANN inputs. These ECN analysis parameters were derived from the time domain, the frequency domain and the time-frequency domain analysis methods. Based on the results of this study, ANN has a very good performance in determining the pitting stage based on ECN analysis results.

Combining Pole Placement and Online Empirical Mode Decomposition Methods to Adaptive Active Control of Structural Vibration

Abstract Using a combination of the pole placement and online empirical mode decomposition (EMD) methods, a new algorithm is proposed for adaptive active control of structural vibration. The EMD method is a time-frequency domain analysis method that can be used for nonstationary and nonlinear problems. Combining the EMD method and Hilbert transform, which is called Hilbert–Huang transform, achieves a method that can be implemented to extract instantaneous properties of signals such as structural response dominant instantaneous frequencies. In the proposed algorithm, these estimated instantaneous properties are utilized to improve the pole-placement method as an adaptive active control technique. The required active control gains are obtained using a genetic algorithm scheme, and optimal gains are calculated corresponding to preselected excitation frequencies. An algorithm is also introduced to choose excitation frequencies based on online EMD method resolution. In order to investigate the efficiency of the proposed method, some case studies that include a discrete model, continuous samples of beam and plate structures, and experimental cantilevered beam are carried out, and the results of the proposed method are compared with the preset (nonadaptive) optimal gains conditions.

Application of the Variational Mode Decomposition-Based Time and Time–Frequency Domain Analysis on Series DC Arc Fault Detection of Photovoltaic Arrays

Series DC arc fault can cause fire hazards in the photovoltaic(PV) array. This paper proposes a time and time-frequency domain analysis method combining the loop current and the PV-side voltage for detecting the series DC arc fault. The fusion of the two different signals can enhance the anti-interference ability of the algorithm. In the time domain analysis, the conductance is put forward to represent the the circuit states. In comparison with some common feature characteristics including current and voltage, the changes of the conductance are more obvious when the arc occurs. In the time-frequency domain analysis, the Variational Mode Decomposition (VMD) is firstly adopted to extract the characteristic frequency band of the arc current signals. VMD can improve the quality of the frequency bands by conquering the modal aliasing and endpoints effects of some traditional modal decomposition algorithms. Then, shannon entropy of the corresponding frequency-band signals is calculated to reflect the variation of the signal complexity. Finally, the two optimal detection variable with rectangle window and the proper time window length are established to achieve the best identification results. In the experimental phase, the experimental results validate that the presented algorithm can not only detect the arc faults timely and accurately but also avoid the nuisance trips caused by the start and shutdown operation of grid-connected inverter with MPPT, the dynamic changes of load and shadow occlusion.

Hydrocarbon identification by application of improved sparse constrained inverse spectral decomposition to frequency-dependent AVO inversion

Spectral decomposition based on sparse constrained inversion, as proposed in recent years, is a high-resolution time-frequency domain analysis method. However, the traditional inverse spectral decomposition (ISD) algorithm is based on an inverse wavelet transform (IWT) whose mother wavelet is a Ricker wavelet function instead of an adaptive wavelet, which causes poor continuity and instability of the ISD results. In this paper, we have extended ISD into other linear transforms and have developed an ISD based on inverse S transform (ISD-IST), which has higher continuity and stability due to the more adjustable and adaptable mother wavelet of the S transform. Comparison of conventional methods and the ISD-IST shows that the ISD (ISD-IWT and ISD-IST) methods have higher resolution and accuracy and can provide more reliable time-frequency phase spectra, and it further verifies the greater continuity and stability of ISD-IST over ISD-IWT. Frequency-dependent AVO (FAVO) inversion is regarded as a potential hydrocarbon identification technique (Wilson et al 2009 SEG Technical Program Expanded Abstracts pp 341–5). However, it is frequency-decomposition based, which means that inversion results can be greatly affected by the effect of time-frequency analysis (Wu et al 2010 SEG Technical Program Expanded Abstracts pp 425–9; Wu et al 2014 Geophys. Prospect. 62 1224–37). Combining the FAVO inversion and high-resolution time-frequency decomposition ISD allows better dispersion attributes to be obtained. Results from application in Southwest China demonstrate that the high frequency-resolution of ISD is helpful in the extraction of gas-induced frequency anomalies from post-stack data, and its high time-resolution helps to obtain high time-resolution gas-induced dispersion results from FAVO inversion. The results of FAVO-ISD not only agree better with well information, but also supply more credible boundary information for the reservoirs. ISD-IST can provide more continuous results and smoother boundary information for the gas reservoir than ISD-IWT.