Showing papers on "Time–frequency analysis published in 2019"
TL;DR: Conv-TasNet as discussed by the authors uses a linear encoder to generate a representation of the speech waveform optimized for separating individual speakers, which is achieved by applying a set of weighting functions masks to the encoder output.
Abstract: Single-channel, speaker-independent speech separation methods have recently seen great progress. However, the accuracy, latency, and computational cost of such methods remain insufficient. The majority of the previous methods have formulated the separation problem through the time–frequency representation of the mixed signal, which has several drawbacks, including the decoupling of the phase and magnitude of the signal, the suboptimality of time–frequency representation for speech separation, and the long latency in calculating the spectrograms. To address these shortcomings, we propose a fully convolutional time-domain audio separation network Conv-TasNet, a deep learning framework for end-to-end time-domain speech separation. Conv-TasNet uses a linear encoder to generate a representation of the speech waveform optimized for separating individual speakers. Speaker separation is achieved by applying a set of weighting functions masks to the encoder output. The modified encoder representations are then inverted back to the waveforms using a linear decoder. The masks are found using a temporal convolutional network consisting of stacked one-dimensional dilated convolutional blocks, which allows the network to model the long-term dependencies of the speech signal while maintaining a small model size. The proposed Conv-TasNet system significantly outperforms previous time–frequency masking methods in separating two- and three-speaker mixtures. Additionally, Conv-TasNet surpasses several ideal time–frequency magnitude masks in two-speaker speech separation as evaluated by both objective distortion measures and subjective quality assessment by human listeners. Finally, Conv-TasNet has a significantly smaller model size and a shorter minimum latency, making it a suitable solution for both offline and real-time speech separation applications. This study, therefore, represents a major step toward the realization of speech separation systems for real-world speech processing technologies.
537 citations
TL;DR: The estimation error shows that the presented algorithm is comparable to the minimum mean square error (MMSE) with full knowledge of the channel statistics, and it is better than an approximation to linear MMSE.
Abstract: In this letter, we present a deep learning algorithm for channel estimation in communication systems. We consider the time–frequency response of a fast fading communication channel as a 2D image. The aim is to find the unknown values of the channel response using some known values at the pilot locations. To this end, a general pipeline using deep image processing techniques, image super-resolution (SR), and image restoration (IR) is proposed. This scheme considers the pilot values, altogether, as a low-resolution image and uses an SR network cascaded with a denoising IR network to estimate the channel. Moreover, the implementation of the proposed pipeline is presented. The estimation error shows that the presented algorithm is comparable to the minimum mean square error (MMSE) with full knowledge of the channel statistics, and it is better than an approximation to linear MMSE. The results confirm that this pipeline can be used efficiently in channel estimation.
373 citations
Journal Article•
TL;DR: The proposed TFA method is based on synchrosqueezing transform and employs an iterative reassignment procedure to concentrate the blurry TF energy in a stepwise manner, meanwhile retaining the signal reconstruction ability.
Abstract: Time-frequency (TF) analysis (TFA) method is an important tool in industrial engineering fields. However, restricted to Heisenberg uncertainty principle or unexpected cross terms, the classical TFA methods often generate blurry TF representation, which heavily hinder its engineering applications. How to generate the concentrated TF representation for a strongly time-varying signal is a challenging task. In this paper, we propose a new TFA method to study the nonstationary features of strongly time-varying signals. The proposed method is based on synchrosqueezing transform and employs an iterative reassignment procedure to concentrate the blurry TF energy in a stepwise manner, meanwhile retaining the signal reconstruction ability. Two implementations of the discrete algorithm are provided, which show that the proposed method has limited computational burden and has potential in real-time application. Moreover, we introduce an effective algorithm to detect the instantaneous frequency trajectory, which can be used to decompose monocomponent modes. Numerical and real-world signals are employed to validate the effectiveness of the proposed method by comparing with some advanced methods. By comparisons, it is shown that the proposed method has the better performance in addressing strongly time-varying signals and noisy signals.
202 citations
TL;DR: Alternative formulations of Morlet wavelets in time and in frequency are presented that allow parameterizing the wavelets directly in terms of the desired temporal and spectral smoothing (expressed as full-width at half-maximum).
160 citations
TL;DR: A CNN-based modulation recognition framework for the detection of radio signals in communication systems and shows that the proposed CNN architecture with spectrogram images as signal representation achieves better recognition accuracy than existing deep learning-based methods.
Abstract: Recent convolutional neural networks (CNNs)-based image processing methods have proven that CNNs are good at extracting features of spatial data. In this letter, we present a CNN-based modulation recognition framework for the detection of radio signals in communication systems. Since the frequency variation with time is the most important distinction among radio signals with different modulation types, we transform 1-D radio signals into spectrogram images using the short-time discrete Fourier transform. Furthermore, we analyze statistical features of the radio signals and use a Gaussian filter to reduce noise. We compare the proposed CNN framework with two existing methods from literature in terms of recognition accuracy and computational complexity. The experiments show that the proposed CNN architecture with spectrogram images as signal representation achieves better recognition accuracy than existing deep learning-based methods.
119 citations
TL;DR: This paper presents a new decomposition approach called adaptive chirp mode pursuit (ACMP), similar to the matching pursuit method, the ACMP captures signal modes one by one in a recursive framework.
116 citations
TL;DR: A joint time-frequency (TF) squeezing method and generalized demodulation (GD) to realize variable speed bearing fault diagnosis and has better performance than those methods based on conventional TF analysis and resampling.
Abstract: High-resolution time-frequency representation (TFR) method is effective for signal analysis and feature detection. However, for variable speed bearing vibration signal, conventional TFR method is prone to blur and affect the accuracy of the instantaneous frequency estimation. Moreover, the traditional order tracking, relying on equi-angular resampling, usually suffers from interpolation error. To solve such problems, we propose a joint time-frequency (TF) squeezing method and generalized demodulation (GD) to realize variable speed bearing fault diagnosis. The method can represent the time-varying fault characteristic frequency precisely and be free from resampling. First, using fast spectral kurtosis to select the optimal-frequency band which is sensitive to rolling bearing fault, and extracting envelope by Hilbert transform within the selected optimal frequency band. Next, a high-quality TF clustering method based on short-time Fourier transform is applied to the TF analysis of the envelope to get a clear TFR, from which the frequency information for GD is obtained. Finally, processing the basic demodulator via the peak search through the TF analysis results in the TFR for GD to gain a resampling-free-order spectrum. Based on the more precise TF information from the clearer TFR, the bearing fault can be diagnosed via GD without tachometer or any resampling involved, avoiding the amplitude error and low computational efficiency of resampling. Simulation study and experimental signal analysis validate that the proposed method has better performance than those methods based on conventional TF analysis and resampling.
104 citations
TL;DR: This work investigates a low complexity linear minimum mean square error receiver which exploits sparsity and quasi-banded structure of matrices involved in the demodulation process which results in a log-linear order of complexity without any performance degradation of BER.
Abstract: Orthogonal time frequency space modulation is a two dimensional (2D) delay-Doppler domain waveform. It uses inverse symplectic Fourier transform (ISFFT) to spread the signal in time-frequency domain. To extract diversity gain from 2D spreaded signal, advanced receivers are required. In this work, we investigate a low complexity linear minimum mean square error receiver which exploits sparsity and quasi-banded structure of matrices involved in the demodulation process which results in a log-linear order of complexity without any performance degradation of BER.
98 citations
TL;DR: This paper proposes a phase-aware speech enhancement algorithm based on DNN to transform an unstructured phase spectrogram to its derivative along the time axis, i.e., instantaneous frequency deviation (IFD), which has a similar structure with its corresponding magnitude spectrogram.
Abstract: Short-time frequency transform (STFT) is fundamental in speech processing Because of the difficulty of processing highly unstructured STFT phase, most speech-processing algorithms only operate with STFT magnitude, leaving the STFT phase far from explored However, with the recent development of deep neural network (DNN) based speech processing, eg, speech enhancement and recognition, phase processing is becoming more important than ever before as a new growing point of DNN-based methods In this paper, we propose a phase-aware speech enhancement algorithm based on DNN Specifically, in the training stage, when incorporating phase as a target, our core idea is to transform an unstructured phase spectrogram to its derivative along the time axis, ie, instantaneous frequency deviation (IFD), which has a similar structure with its corresponding magnitude spectrogram We further propose to optimize both IFD and magnitude jointly in a multiobjective learning framework In the test stage, we propose a postprocessing method to recover the phase spectrogram from the estimated IFD Experimental results demonstrate the effectiveness of the proposed method
75 citations
TL;DR: A new SST method called high-order synchrosqueezing wavelet transform is proposed to achieve a highly energy-concentrated time-frequency representation (TFR) for nonstationary signals with wide frequency range and fast varying IF.
71 citations
Posted Content•
TL;DR: In this article, a low complexity linear minimum mean square error (MLMSE) receiver is proposed to exploit sparsity and quasi-banded structure of matrices involved in the demodulation process, which results in a loglinear order of complexity without any performance degradation of BER.
Abstract: Orthogonal time frequency space modulation is a two dimensional (2D) delay-Doppler domain waveform. It uses inverse symplectic Fourier transform (ISFFT) to spread the signal in time-frequency domain. To extract diversity gain from 2D spreaded signal, advanced receivers are required. In this work, we investigate a low complexity linear minimum mean square error receiver which exploits sparsity and quasi-banded structure of matrices involved in the demodulation process which results in a log-linear order of complexity without any performance degradation of BER.
TL;DR: A novel SST-based technique is proposed that can achieve more concentrated representations than RM and SST and allows for perfect signal reconstruction and the reconstructed signal has a high consistency with the general relativity proposed by Einstein.
TL;DR: This paper proposes to set the parameters of the GST adaptively using the instantaneous frequency (IF) of seismic traces, and names the proposed SAGST as the self-adaptive GST (SAGST).
Abstract: Achieving a proper time–frequency (TF) resolution is the key to extract information from seismic data using TF algorithms and characterize reservoir properties using decomposed frequency components. The generalized S-transform (GST) is one of the most widely used TF algorithms. However, it is difficult to choose an optimized parameter set for the whole seismic data set. In this paper, we propose to set the parameters of the GST adaptively using the instantaneous frequency (IF) of seismic traces. Our workflow begins with building a relationship between the parameter set of the GST and IF using a synthetic wedge model. We use the IF as an indicator for the time thickness of each trace in the wedge model. We then compute the TF spectrum of each trace using the GST with different parameter sets and compare the similarity between the computed TF spectrum and theory TF spectrum. The parameter set with the largest similarity is regarded as the best parameter set for each trace in the wedge model. In this manner, we build a relationship between the parameter set and IF value. We can finally choose the optimum parameter set for the GST according to the IF values of seismic traces. We name the proposed workflow as the self-adaptive GST (SAGST). To demonstrate the validity and effectiveness of the proposed SAGST, we apply it to synthetic seismic traces and field data. Both synthetic and real data examples illustrate that the SAGST can obtain a TF representation with a high TF resolution.
TL;DR: Case studies and comparisons with the continuous Morlet wavelet transform (CMWT) and the tunable Q-factor wavelettransform (TQWT) demonstrate the effectiveness and superiority of the CMQGWT for bearing diagnostic information extraction and fault identification.
Abstract: Rolling element bearings are key and also vulnerable machine elements in rotating machinery. Fault diagnosis of rolling element bearings is significant for guaranteeing machinery safety and functionality. To accurately extract bearing diagnostic information, a time-frequency analysis method based on continuous wavelet transform (CWT) and multiple Q-factor Gabor wavelets (MQGWs) (termed CMQGWT) is introduced in this paper. In the CMQGWT method, Gabor wavelets with multiple Q-factors are adopted and sets of the continuous wavelet coefficients for each Q-factor are combined to generate time-frequency map. By this way, the resolution of the CWT time-frequency map can be greatly increased and the diagnostic information can be accurately identified. Numerical simulation is carried out and verified the effectiveness of the proposed method. Case studies and comparisons with the continuous Morlet wavelet transform (CMWT) and the tunable Q-factor wavelet transform (TQWT) demonstrate the effectiveness and superiority of the CMQGWT for bearing diagnostic information extraction and fault identification.
TL;DR: The proposed time-frequency analysis method is further extended to generate time-varying amplitude and frequency demodulated spectra, inspired by the fact that gear fault frequency is manifested straight by the amplitude andfrequency modulating frequencies.
TL;DR: The results show that the proposed DHST method is more effective in processing the nonstationary signals with fast varying instantaneous frequency than the proposed TFA method.
Abstract: Time–frequency analysis (TFA) is considered as a useful tool to extract the time-variant features of the nonstationary signal. In this paper, a new method called demodulated high-order synchrosqueezing transform (DHST) is proposed. The DHST introduces a two-step algorithm, namely, demodulated transform and high-order synchrosqueezing method to achieve a compact time–frequency representation (TFR) while enabling the reconstruction of the signal from TFR. The performance of the proposed DHST method in this paper is validated by both the simulated and experimental signals including bat echolocation and a vibration signal. The results show that the proposed TFA method is more effective in processing the nonstationary signals with fast varying instantaneous frequency.
TL;DR: The proposed VSLCT is an extended version of the current linear transform that can effectively alleviate the smear effect and can dynamically provide desirable time–frequency resolution in response to condition variations.
Abstract: Linear transform has been widely used in time–frequency analysis of rotational machine vibration. However, the linear transform and its variants in current forms cannot be used to reliably analyze rotational machinery vibration signals under nonstationary conditions because of their smear effect and limited time variability in time–frequency resolution. As such, this paper proposes a new time–frequency method, named velocity synchronous linear chirplet transform (VSLCT). The proposed VSLCT is an extended version of the current linear transform. It can effectively alleviate the smear effect and can dynamically provide desirable time–frequency resolution in response to condition variations. The smearing problem is resolved by using linear chirplet bases with frequencies synchronous with shaft rotational velocity, and the time–frequency resolution is made responsive to signal condition changes using time-varying window lengths. To successfully implement the VSLCT, a kurtosis-guided approach is proposed to dynamically determine the two time-varying parameters, i.e., window length and normalized angle. Therefore, the VSLCT does not require the user to provide such parameters and hence avoids the subjectivity and bias of human judgment that is often time-consuming and knowledge-demanding. This method can also analyze normal monocomponent frequency-modulated signal.
TL;DR: This paper presents a time–frequency masking based online multi-channel speech enhancement approach that uses a convolutional recurrent neural network to estimate the mask and demonstrates the robustness of the system to different angular positions of the speech source.
Abstract: This paper presents a time–frequency masking based online multi-channel speech enhancement approach that uses a convolutional recurrent neural network to estimate the mask. The magnitude and phase components of the short-time Fourier transform coefficients for multiple time frames are provided as an input such that the network is able to discriminate between the directional speech and the noise components based on the spatial characteristics of the individual signals as well as their spectro-temporal structure. The estimation of two different masks, namely, ideal ratio mask (IRM) and ideal binary mask (IBM), along with two different approaches for incorporating the mask to obtain the desired signal are discussed. In the first approach, the mask is directly applied as a real valued gain to a reference microphone signal, whereas in the second approach, the masks are used as an activity indicator for the recursive update of power spectral density (PSD) matrices to be used within a beamformer. The performance of the proposed system with the two different estimated masks utilized within the two different enhancement approaches is evaluated with both simulated as well as measured room impulse responses, where it is shown that the IBM is better suited as an indicator for the PSD updates while direct application of IRM as a real valued gain leads to a better improvement in terms of short term objective intelligibility. Analysis of the performance of the proposed system also demonstrates the robustness of the system to different angular positions of the speech source.
TL;DR: The proposed FTFM-based reconstruction method indicates attractive prospects in the following two aspects: effective but efficient TFM learning for practical and on-line application, sound and adaptive signal reconstruction with the data-driven F TFM basis.
TL;DR: A new time-frequency based PAC (t-f PAC) measure is proposed that is more robust to varying signal parameters and provides a more accurate measure of coupling strength.
Abstract: Oscillatory activity in the brain has been associated with a wide variety of cognitive processes including decision making, feedback processing, and working memory. The high temporal resolution provided by electroencephalography (EEG) enables the study of variation of oscillatory power and coupling across time. Various forms of neural synchrony across frequency bands have been suggested as the mechanism underlying neural binding. Recently, a considerable amount of work has focused on phase-amplitude coupling (PAC)- a form of cross-frequency coupling where the amplitude of a high frequency signal is modulated by the phase of low frequency oscillations. The existing methods for assessing PAC have some limitations including limited frequency resolution and sensitivity to noise, data length and sampling rate due to the inherent dependence on bandpass filtering. In this paper, we propose a new time-frequency based PAC (t-f PAC) measure that can address these issues. The proposed method relies on a complex time-frequency distribution, known as the Reduced Interference Distribution (RID)-Rihaczek distribution, to estimate both the phase and the envelope of low and high frequency oscillations, respectively. As such, it does not rely on bandpass filtering and possesses some of the desirable properties of time-frequency distributions such as high frequency resolution. The proposed technique is first evaluated for simulated data and then applied to an EEG speeded reaction task dataset. The results illustrate that the proposed time-frequency based PAC is more robust to varying signal parameters and provides a more accurate measure of coupling strength.
TL;DR: In this article, the authors show that there exist fast constructions for computing approximate projections onto the leading Slepian basis elements of the discrete Prolate Spheroidal Sequence (DPSS).
TL;DR: In this article, the authors proposed a method for the detection of broken rotor bars in induction machines based on the fact that the fault-related harmonics will have oscillating amplitudes due to the speed ripple effect.
Abstract: This paper offers a reliable solution to the detection of broken rotor bars in induction machines with a novel methodology, which is based on the fact that the fault-related harmonics will have oscillating amplitudes due to the speed ripple effect. The method consists of two main steps: Initially, a time–frequency transformation is used and the focus is given on the steady-state regime; thereupon, the fault-related frequencies are handled as periodical signals over time and the classical fast Fourier transform is used for the evaluation of their own spectral content. This leads to the discrimination of subcomponents related to the fault and to the evaluation of their amplitudes. The versatility of the proposed method relies on the fact that it reveals the aforementioned signatures to detect the fault, regardless of the spatial location of the broken rotor bars. Extensive finite element simulations on a 1.1 MW induction motor and experimental testing on a 1.1 kW induction motor lead to the conclusion that the method can be generalized on any type of induction motor independently from the size, power, number of poles, and rotor slot numbers.
TL;DR: The joint time-frequency scattering transform (JTF) as mentioned in this paper is a time-shift invariant representation that characterizes the multiscale energy distribution of a signal in time and frequency.
Abstract: In time series classification and regression, signals are typically mapped into some intermediate representation used for constructing models. Since the underlying task is often insensitive to time shifts, these representations are required to be time-shift invariant. We introduce the joint time–frequency scattering transform, a time-shift invariant representation that characterizes the multiscale energy distribution of a signal in time and frequency. It is computed through wavelet convolutions and modulus non-linearities and may, therefore, be implemented as a deep convolutional neural network whose filters are not learned but calculated from wavelets. We consider the progression from mel-spectrograms to time scattering and joint time–frequency scattering transforms, illustrating the relationship between increased discriminability and refinements of convolutional network architectures. The suitability of the joint time–frequency scattering transform for time-shift invariant characterization of time series is demonstrated through applications to chirp signals and audio synthesis experiments. The proposed transform also obtains state-of-the-art results on several audio classification tasks, outperforming time scattering transforms and achieving accuracies comparable to those of fully learned networks.
01 Mar 2019-Precision Engineering-journal of The International Societies for Precision Engineering and Nanotechnology
TL;DR: A novel feature extraction approach for chatter detection by using image analysis of dominant frequency bands from the short-time Fourier transform (STFT) spectrograms to indicate the efficiency of the time-frequency image features from dominant Frequency bands for chatter Detection and their better performance than the time domain features and wavelet-based features in terms of their separability capabilities.
Abstract: Chatter is a cause of low surface quality and productivity in milling and crucial features need to be extracted for accurate chatter detection and suppression. This paper introduces a novel feature extraction approach for chatter detection by using image analysis of dominant frequency bands from the short-time Fourier transform (STFT) spectrograms. In order to remove the environmental noises and highlight chatter related characteristics, dominant frequency bands with high energy are identified by applying the squared energy operator to the synthesized fast Fourier transform (FFT) spectrum. The time-frequency spectrogram of the vibration signal is divided into a set of grayscale sub-images according to the dominant frequency bands. Statistical image features are extracted from those sub-images to describe the machining condition and assessed in terms of their separability capabilities. The proposed feature extraction method is verified by using dry milling tests of titanium alloy Ti6Al4V and compared with two existing feature extraction techniques. The results indicate the efficiency of the time-frequency image features from dominant frequency bands for chatter detection and their better performance than the time domain features and wavelet-based features in terms of their separability capabilities.
TL;DR: In this article, a parameterized time-frequency transform (PTFT) method is proposed to estimate the instantaneous rotation frequency (IRF) from vibration signals directly and a maximum correlated kurtosis deconvolution (MCKD) based envelope order spectrum is applied to detect the bearing fault characteristic order.
Abstract: Rolling bearing vibration signals induced by local faults are highly correlated with structural dynamics. This leads to a great deal of research about signal processing for rolling bearing condition monitoring and fault detection. However, the common diagnostic approaches which were proposed to manage vibration signals with constant speeds are unavailable for variable speed cases. Tacholess order tracking is a powerful method to break the limitation of conventional methods while avoiding trouble with tachometer installation and reducing measurement cost. Time frequency analysis methods are used to estimate the instantaneous rotation frequency (IRF) from vibration signals directly. However, it is difficult to extract IRF accurately due to the strong nonstationary property of the signal. Therefore, parameterized time-frequency transform (PTFT) methods are proposed to solve this problem. Polynomial chirplet transform (PCT) is one of the PTFTs that can produce an excellent time frequency representation (TFR) with a polynomial kernel function. In this paper, the PCT is employed to estimate the IRF of rolling bearings from the vibration signals. On this basis, a maximum correlated kurtosis deconvolution (MCKD) based envelope order spectrum is applied to detect the bearing fault characteristic order (FCO). The efficiency of the proposed method is certified by numerical signal and rolling bearing vibration data. The diagnostic results indicate that the new fault detection algorithm is superior for rolling bearing fault diagnosis under varying speed conditions.
TL;DR: The Short Time Fourier Transform (STFT) is proposed in this paper; giving additional information on changes of the frequencies over time for stator current signal analysis.
Abstract: Induction motor diagnosis using the Power Spectral Density (PSD) estimation based on the Fourier Transform calculation has been widely used as an analysis method for its simplicity and low computation time. However, the use of PSD is not recommended for processing non stationary signals (case of variable speed applications) and therefore the analysis with PSD is not reliable. To overcome this handicap, the Short Time Fourier Transform (STFT) is proposed in this paper; giving additional information on changes of the frequencies over time for stator current signal analysis. Furthermore, the use of a new approach called Maxima’s Location Algorithm is also proposed. This later will be associated with the STFT analysis to show only those harmonics with useful information on existing faults. This approach will be used in the diagnosis of bearing faults of a PWM inverter-fed induction motor operating at variable speed. Several experimental results in the transient state are carried out firstly to validate the results and secondly to illustrate the merits and effectiveness of the combined STFT/MLA proposed approach.
TL;DR: This paper reviews approaches used to detect and identify arcing currents, including arcing current faults, categorized as the time-domain, frequency- domain, and time–frequency approaches.
Abstract: This paper reviews approaches used to detect and identify arcing currents, including arcing current faults. The reviewed approaches are categorized as the time-domain, frequency-domain, and time–frequency approaches. The time-domain approach extracts shoulders (zero values of the current around zero crossing points), spikes and jumps, abnormal magnitudes (lower or higher than normal), and high rate of change of the current. The frequency-domain approach extracts the high-frequency components, harmonic components, sub-harmonic components, and cross-correlation indicator. The time–frequency approach extracts high-frequency sub-bands that contain non-stationary frequency components, which may have non-stationary phases. The three approaches are implemented to test their accuracy, computational requirements, and sensitivity to system parameters. These tests are performed by processing of currents that are collected for normal and dynamic conditions, conventional faults, and currents with high or low arcing components. Test results provide a performance comparison for the tested approaches.
TL;DR: This paper addresses the problem of constructing a feature extractor which combines Mallat's scattering transform framework with time-frequency (Gabor) representations, and introduces a class of frames, called uniform covering frames, which includes a variety of semi-discrete Gabor systems.
TL;DR: The concept of synchrosqueezing transforms (SSTs) that was developed to sharpen linear time–frequency representations (TFRs) in such a way that the sharpened transforms remain invertible are introduced.
TL;DR: Experimental results indicate that the proposed method outperforms state-of-the-art methods such as original Viterbi-based IF estimation algorithm and ridge path regrouping methods.
Abstract: Time–frequency (TF)-based instantaneous frequency estimation algorithms fail to achieve the desired performance when the underlying TF distribution suffers from low resolution of the signal components or signal components intersect each other in the TF domain. This paper addresses above-mentioned problems by (a) employing adaptive directional time–frequency distributions for resolving close components and (b) developing a variant of the Viterbi algorithm that employs both the direction and amplitude of the signal components for IF estimation of crossing components at low signal-to-noise ratio. Experimental results indicate that the proposed method outperforms state-of-the-art methods such as original Viterbi-based IF estimation algorithm and ridge path regrouping methods.