Time-frequency Feature Extraction Research Articles

Continuous Wavelet Transform for Decoding Finger Movements From Single-Channel EEG.

Human body movements can be reflected in brain signals and collected noninvasively with electroencephalography (EEG). Motor-related signals include sensory motor rhythms (also known as the Mu wave) in the upper-alpha band of 8-13Hz and slow cortical potentials (SCPs) in the low frequency range of 0.1-5Hz. This study compares the two signals for decoding finger movements. Human subjects were asked to individually lift each of the five digits of their right hand, at the rate of one every 10 s. EEG was recorded using a bipolar montage between ipsilateral and contralateral motor cortices. Electromyograms were obtained for identifying movement onsets. Linear discriminant analysis (LDA) generated significant performance with SCPs but not with Mu. Meanwhile, continuous wavelet transform (CWT) was applied to SCPs or Mu to create a spectrogram for each finger, showing distinctive amplitude and phase patterns. A dprime-based weighting algorithm was used to extract time-frequency features. With a template-matching paradigm, both SCP and Mu spectrograms yielded significant classification accuracies for multiple subjects, with the highest being >50% (chance = 20%). Interestingly, the index finger was better distinguished with Mu for most of the subjects, whereas the ring finger was better distinguished with SCPs. The CWT algorithm outperformed LDA by better decoding the thumb. This study suggests that the time-frequency characteristics of a single-channel EEG, when phase is preserved, contain critical information on finger movements. SCPs and Mu seem to work in an independent but complementary manner.

EEG-Based Emotion Recognition Using Quadratic Time-Frequency Distribution.

Accurate recognition and understating of human emotions is an essential skill that can improve the collaboration between humans and machines. In this vein, electroencephalogram (EEG)-based emotion recognition is considered an active research field with challenging issues regarding the analyses of the nonstationary EEG signals and the extraction of salient features that can be used to achieve accurate emotion recognition. In this paper, an EEG-based emotion recognition approach with a novel time-frequency feature extraction technique is presented. In particular, a quadratic time-frequency distribution (QTFD) is employed to construct a high resolution time-frequency representation of the EEG signals and capture the spectral variations of the EEG signals over time. To reduce the dimensionality of the constructed QTFD-based representation, a set of 13 time- and frequency-domain features is extended to the joint time-frequency-domain and employed to quantify the QTFD-based time-frequency representation of the EEG signals. Moreover, to describe different emotion classes, we have utilized the 2D arousal-valence plane to develop four emotion labeling schemes of the EEG signals, such that each emotion labeling scheme defines a set of emotion classes. The extracted time-frequency features are used to construct a set of subject-specific support vector machine classifiers to classify the EEG signals of each subject into the different emotion classes that are defined using each of the four emotion labeling schemes. The performance of the proposed approach is evaluated using a publicly available EEG dataset, namely the DEAPdataset. Moreover, we design three performance evaluation analyses, namely the channel-based analysis, feature-based analysis and neutral class exclusion analysis, to quantify the effects of utilizing different groups of EEG channels that cover various regions in the brain, reducing the dimensionality of the extracted time-frequency features and excluding the EEG signals that correspond to the neutral class, on the capability of the proposed approach to discriminate between different emotion classes. The results reported in the current study demonstrate the efficacy of the proposed QTFD-based approach in recognizing different emotion classes. In particular, the average classification accuracies obtained in differentiating between the various emotion classes defined using each of the four emotion labeling schemes are within the range of –. Moreover, the emotion classification accuracies achieved by our proposed approach are higher than the results reported in several existing state-of-the-art EEG-based emotion recognition studies.

Dynamic Time-frequency Feature Extraction for Brain Activity Recognition.

The biomedical signal classification accuracy on motor imagery is not always satisfactory, partially because not all the important features have been effectively extracted. This paper proposes an improved dynamic feature extraction approach based on a time-frequency representation and an optimal sequence similarity measurement. Since the wavelet packet decomposition (WPD) generates more detailed signal variation information and the dynamic time warping (DTW) helps optimally measure the sequence similarity, more important features are kept for classification. We apply the extracted features from our proposed method to Electroencephalogram (EEG) based motor imagery through the OpenBCI device and obtain higher classification accuracy. Compared with traditional feature extraction methods, there is a significant classification accuracy improvement from 83.53% to 90.89%. Our work demonstrates the importance of the advanced feature extraction in time series data analysis, e.g. biomedical signal.

FallDeFi

Falling or tripping among elderly people living on their own is recognized as a major public health worry that can even lead to death. Fall detection systems that alert caregivers, family members or neighbours can potentially save lives. In the past decade, an extensive amount of research has been carried out to develop fall detection systems based on a range of different detection approaches, i.e, wearable and non-wearable sensing and detection technologies. In this paper, we consider an emerging non-wearable fall detection approach based on WiFi Channel State Information (CSI). Previous CSI based fall detection solutions have considered only time domain approaches. Here, we take an altogether different direction, time-frequency analysis as used in radar fall detection. We use the conventional Short-Time Fourier Transform (STFT) to extract time-frequency features and a sequential forward selection algorithm to single out features that are resilient to environment changes while maintaining a higher fall detection rate. When our system is pre-trained, it has a 93% accuracy and compared to RTFall and CARM, this is a 12% and 15% improvement respectively. When the environment changes, our system still has an average accuracy close to 80% which is more than a 20% to 30% and 5% to 15% improvement respectively.

Epileptic electroencephalogram recognition based on discrete S transform and permutation entropy

Electroencephalogram(EEG) analysis has important reference value in the diagnosis of epilepsy. The automatic classification of epileptic EEG can be used to judge the patient's situation in time,which is of great significance in clinical application. In order to solve the problem that the recognition accuracy is not high by using the single feature of EEG signals and avoid the influence of wavelet basis function selection on recognition results,a method of automatic discrimination of epileptic EEG signals based on S transform and permutation entropy is proposed. Firstly, the original signals are decomposed by discrete S transform, and then we calculate the fluctuation index of coefficients of each rhythm and combine the permutation entropy of EEG signals into a feature vector and use Real AdaBoost classifier to discriminate the epileptic EEG signals in muti-period. In this study, we used the epilepsy database from University of Bonn. Three groups of EEG signals, including the data of normal people with their eyes open, the data collected inside of the epileptic foci from patients during their interictal period and the data during their ictal period, were used to test effectiveness. The results of this study showed that the fluctuation index of each rhythm could be used to characterize the normal, interictal and ictal epileptic EEG signals effectively, and the recognition accuracy of multiple features was much higher than that of single feature. The average recognition accuracy could reach 98.13%. Compared with time-frequency feature extraction method or nonlinear feature extraction method only,the recognition accuracy was increased by more than 1.2% and 8.1% respectively, which was superior to the methods recorded in many other literatures. Therefore, this method has a good application prospect in diagnosis of epilepsy.

On time-frequency domain feature extraction of wave signals for structural health monitoring

Wave propagation signals are commonly used as information carrier in structural health monitoring. To facilitate decision making, wave signals are often decomposed into multiple components to reveal its frequency or time-frequency content. In this research we investigate the use of time-frequency decomposition techniques for feature extraction. The method based on the adaptive harmonic wavelet transform (AHWT) possesses high computational efficiency and at the same time inherently avoids certain issues in some other time-frequency feature extraction methods, e.g., the empirical mode decomposition (EMD). EMD entertains the adaptivity with respect to signal features, but that adaptive nature affects the comparability of the resulting intrinsic mode functions (IMFs). In contrast, the AHWT based method has similar feature extraction capabilities in the time-frequency domain while using a deterministic basis. Therefore, wavelet features can become cross-comparable when common wavelet basis is used. Our case study shows that the AHWT based approach can identify the critical features in Lamb wave signals to realize effective and robust decision making in structural damage detection.

Myoelectric signal classification based on S transform and two-directional two-dimensional principal component analysis

Time-frequency representiation has been intensively employed for the analysis of biomedical signals. In order to extract discriminative information, time-frequency matrix is often transformed into a 1D vector followed by principal component analysis (PCA). This study contributes a two-directional two-dimensional principal component analysis (2D2PCA)-based technique for time-frequency feature extraction. The S transform, integrating the strengths of short time Fourier transform and wavelet transform, is applied to perform the time-frequency decomposition. Then, 2D2PCA is directly conducted on the time-frequency matrix rather than 1D vectors for feature extraction. The proposed method can significantly reduce the computational cost while capture the directions of maximal time-frequency matrix variance. The efficiency and effectiveness of the proposed method is demonstrated by classifying eight hand motions using 4-channel myoelectric signals recorded in health subjects and amputees.

Epileptic Seizure Classification of EEGs Using Time-Frequency Analysis Based Multiscale Radial Basis Functions.

The automatic detection of epileptic seizures from electroencephalography (EEG) signals is crucial for the localization and classification of epileptic seizure activity. However, seizure processes are typically dynamic and nonstationary, and thus, distinguishing rhythmic discharges from nonstationary processes is one of the challenging problems. In this paper, an adaptive and localized time-frequency representation in EEG signals is proposed by means of multiscale radial basis functions (MRBF) and a modified particle swarm optimization (MPSO) to improve both time and frequency resolution simultaneously, which is a novel MRBF-MPSO framework of the time-frequency feature extraction for epileptic EEG signals. The dimensionality of extracted features can be greatly reduced by the principle component analysis algorithm before the most discriminative features selected are fed into a support vector machine (SVM) classifier with the radial basis function (RBF) in order to separate epileptic seizure from seizure-free EEG signals. The classification performance of the proposed method has been evaluated by using several state-of-art feature extraction algorithms and other five different classifiers like linear discriminant analysis, and logistic regression. The experimental results indicate that the proposed MRBF-MPSO-SVM classification method outperforms competing techniques in terms of classification accuracy, and shows the effectiveness of the proposed method for classification of seizure epochs and seizure-free epochs.

A reassigned bilinear transformation and Gaussian kernel function-based approach for detecting weak targets in sea clutter

ABSTRACTThis article proposes an efficient approach for weak target detection in sea clutter by using reassigned bilinear transformation based on the Gaussian kernel function (KF). This approach, which makes no assumptions on the sea clutter, mainly incorporates three steps: time-frequency analysis, feature extraction, and pattern classification. Bilinear transformation is utilized to transform sea clutter data into a time-frequency image, and the suitable Gaussian KF is applied to restrain the cross-term interference. Subsequently, to obtain a better performance in focus-ability, readability, and time-frequency resolution, we reassign the time-frequency distribution. In addition, we employ its marginal distribution of frequency (MARF) as classification feature vectors to decrease the dimensions of time-frequency features. Finally, a multilayered feed-forward neural network (NN) is selected as a classifier in the pattern classification process. Detection performances using the IPIX radar experimental sea clutter data are compared among the proposed method, the method based on continuous wavelet transform, and the conventional constant false alarm rate (CFAR) method based on spectral properties. A comparison of the results shows that the proposed method is more effective and efficient than the other two methods for weak target detection in sea clutter.

Superiorities of variational mode decomposition over empirical mode decomposition particularly in time–frequency feature extraction and wind turbine condition monitoring

Due to constantly varying wind speed, wind turbine (WT) components often operate at variable speeds in order to capture more energy from wind. As a consequence, WT condition monitoring (CM) signals always contain intra-wave features, which are difficult to extract through performing conventional time–frequency analysis (TFA) because none of which is locally adaptive. So far, only empirical mode decomposition (EMD) and its extension forms can extract intra-wave features. However, the EMD and those EMD-based techniques also suffer a number of defects in TFA (e.g. weak robustness of against noise, unidentified ripples, inefficiency in detecting side-band frequencies etc.). The existence of these issues has significantly limited the extensive application of the EMD family techniques to WT CM. Recently, an alternative TFA method, namely variational mode decomposition (VMD), was proposed to overcome all these issues. The purpose of this study is to verify the superiorities of the VMD over the EMD and investigate its potential application to the future WT CM. Experiment has shown that the VMD outperforms the EMD not only in noise robustness but also in multi-component signal decomposition, side-band detection, and intra-wave feature extraction. Thus, it has potential as a promising technique for WT CM.

Automatic signal abnormality detection using time-frequency features and machine learning: A newborn EEG seizure case study

Time-frequency (TF) based machine learning methodologies can improve the design of classification systems for non-stationary signals. Using selected TF distributions (TFDs), TF feature extraction is performed on multi-channel recordings using channel fusion and feature fusion approaches. Following the findings of previous studies, a TF feature set is defined to include three complementary categories: signal related features, statistical features and image features. Multi-class strategies are then used to improve the classification algorithm robustness to artifacts. The optimal subset of TF features is selected using the wrapper method with sequential forward feature selection (SFFS). In addition, a new proposed measure for TF feature selection is shown to improve the sensitivity of the classifier (while slightly reducing total accuracy and specificity). As an illustration, the TF approach is applied to the design of a system for detection of seizure activity in real newborn EEG signals. Experimental results indicate that: (1) The compact kernel distribution (CKD) outperforms other TFDs in classification accuracy; (2) a feature fusion strategy gives better classification than a channel fusion strategy; e.g. using all TF features, the CKD achieves a classification accuracy of 82% with feature fusion, which is 4% more than the channel fusion approach; (3) the SFFS wrapper feature selection method improves the classification performance for all TFDs; e.g. total accuracy is improved by 4.6%; (4) the multi-class strategy improves the seizure detection accuracy in the presence of artifacts; e.g. a total accuracy of 86.61% with one vs. one multi-class approach is achieved i.e. 0.91% more than the binary classification approach. The results obtained on a large practical real data set confirm the improved performance capability of TF features for knowledge based systems.

Activity recognition and intensity estimation in youth from accelerometer data aided by machine learning

Physical activity monitoring for youth is an area of increasing scientific and public health interest due to the high prevalence of obesity and downward trend in physical activity. However, accurate assessment of such activity remains a challenging problem because of the complex nature in which certain activities are performed. In this study, we formulated the issue as a machine learning problem—using a diverse set of 19 physical activities commonly performed by youth—via two approaches: activity recognition and intensity estimation. With the aid of training data, we implemented a distance metric learning method called DML-KNN that utilizes time-frequency features and is capable of effectively classifying both continuous and intermittent movement in youth subjects. Four different time-frequency feature extraction methods were then systematically evaluated. Our results show that the DML-KNN method performed competitively, especially when using features extracted by the Tamura method for intensity estimation, and by the Square Coefficient method for activity recognition.

Characterizing slug to churn flow transition by using multivariate pseudo Wigner distribution and multivariate multiscale entropy

Uncovering transitional behavior of slug flow under different flow conditions constitutes a challenging problem of significant importance. Based on the experimental multi-channel measurements, we first calculate multivariate pseudo Wigner distribution (MPWD) and extract time–frequency features to probe the transient behavior of slug flow and churn flow. Then we calculate the multivariate multiscale sample entropy (MMSE) for different flow conditions and find that the MMSE allows us to uncover the dynamic flow behavior governing the transition from slug flow to churn flow. Our results indicate combining the MPWD and MMSE enables to reveal the transient and multiscale flow behavior underlying the formation and transition of slug flow and churn flow.

An Investigation into Error Source Identification of Machine Tools Based on Time-Frequency Feature Extraction

This paper presents a new identification method to identify the main errors of the machine tool in time-frequency domain. The low- and high-frequency signals of the workpiece surface are decomposed based on the Daubechies wavelet transform. With power spectral density analysis, the main features of the high-frequency signal corresponding to the imbalance of the spindle system are extracted from the surface topography of the workpiece in the frequency domain. With the cross-correlation analysis method, the relationship between the guideway error of the machine tool and the low-frequency signal of the surface topography is calculated in the time domain.

Time-Frequency Feature Extraction of HRRP Using AGR and NMF for SAR ATR

A new approach to classify synthetic aperture radar (SAR) targets based on high range resolution profiles (HRRPs) is presented. Features from each of the target HRRPs are extracted via the nonnegative matrix factorization (NMF) algorithm in time-frequency domain represented by adaptive Gaussian representation (AGR). Firstly, SAR target images have been converted into HRRPs. And the time-frequency matrix for each of HRRPs is obtained by using AGR. Secondly, the time-frequency feature vectors are extracted from the time-frequency matrix utilizing NMF. Finally, hidden Markov models (HMMs) are employed to characterize the time-frequency feature vectors corresponding to one target and are used to being the recognizer. To demonstrate the performance of the proposed approach, experiments are performed in the 10-target MSTAR public dataset. The results support the effectiveness of the proposed technique for SAR automatic target recognition (ATR).

Vibration signal analysis using parameterized time–frequency method for features extraction of varying-speed rotary machinery

In real application, when rotary machinery frequently involves variable-speed, unsteady load and defect, it will produce non-stationary vibration signal. Such signal can be characterized by mono- or multi-component frequency modulation (FM) and its internal instantaneous patterns are closely related to operation condition of the rotary machinery. For example, instantaneous frequency (IF) and instantaneous amplitude (IA) of a non-stationary signal are two important time–frequency features to be inspected. For vibration signal analysis of the rotary machinery, time–frequency analysis (TFA), known for analyzing the signal in the time and frequency domain simultaneously, has been accepted as a key signal processing tool. Particularly, parameterized TFA, among various TFAs, has shown great potential to investigate time–frequency features of non-stationary signals. It attracts more attention for improving time–frequency representation (TFR) with signal-dependent transform parameters. However, the parameter estimation and component separation are two problems to tackle with while using the parameterized TFA to extract time–frequency features from non-stationary vibration signal of varying-speed rotary machinery. In this paper, we propose a procedure for the parameterized TFA to analyze the non-stationary vibration signal of varying-speed rotary machinery. It basically includes four steps: initialization, estimation of transform parameter, component separation and parameterized TFA, as well as feature extraction. To demonstrate the effectiveness of the proposed method in analyzing mono- and multi-component signals, it is first used to analyze the vibration response of a laboratory rotor during a speed-up and run-down process, and then extract the instantaneous time–frequency signatures of a hydro-turbine rotor in a hydroelectric power station during a shut-down stage. In addition, the results are compared with several traditional TFAs and the proposed method outperforms others in accurate feature extraction, which is promising in applications of fault detection, system condition monitoring, parameter identification, etc.

An Improved EMD Method for Time–Frequency Feature Extraction of Telemetry Vibration Signal Based on Multi-Scale Median Filtering

We hereby propose an Empirical Mode Decomposition (EMD) method improved with a multi-scale median filtering for extraction of the time---frequency feature of telemetry vibration signals under interference from impulse noise. The signal is decomposed into a series of intrinsic mode functions (IMF) by EMD roughly. Median filtering is then performed on each IMF with filter window length varying with the IMF's frequency, respectively. This maneuver will allow effective impulse noise suppression with minimal loss of signal integrity. A new signal can then be reconstructed by adding up each component after the median filtering and treated with a repeat EMD to obtain new IMFs as a final result. This method overcomes the filtering window length selection problem in the median filtering, which can obtain better time---frequency feature extraction performance under the impulse noise interference condition. Data processing results from both a simulation signal and a telemetry vibration signal of a test showed the effectiveness of this method.

Principles of time–frequency feature extraction for change detection in non-stationary signals: Applications to newborn EEG abnormality detection

This paper considers the general problem of detecting change in non-stationary signals using features observed in the time–frequency (t,f) domain, obtained using a class of quadratic time–frequency distributions (QTFDs). The focus of this study is to propose a methodology to define new (t,f) features by extending time-only and frequency-only features to the joint (t,f) domain for detecting changes in non-stationary signals. The (t,f) features are used as a representative subset characterizing the status of the observed non-stationary signal. Change in the signal is then reflected as a change in the (t,f) features. This (t,f) approach is applied to the problem of detecting abnormal brain activity in newborns (e.g. seizure) using measurements of the EEG for diagnosis and prognosis. In addition, a pre-processing stage for detecting artifacts in EEG signals for signal enhancement is studied and implemented separately. Overall results indicate that, in general, the (t,f) approach results in an improved performance in detecting artifacts and seizures in newborn EEG signals as compared to time-only or frequency-only features.

An analysis of the simulated acoustic emission sources with different propagation distances, types and depths for rail defect detection

Acoustic Emission (AE) technique is an effective nondestructive detecting method, and has a promising application for rail defect detection. So far, little attention has been paid to propagation distances, types, and depths of AE sources, which are important for rail defect detection accurately. This paper presents an experimental study on the simulated AE sources with different propagation distances, types and depths for rail defect detection. Three simulated AE sources with different frequencies are seeded on the cross section of rail, and the depths of AE sources are changed in the vertical direction. After receiving AE signals, wavelet transform and Rayleigh–Lamb equations are utilized to extract time–frequency features and modes. Based on the wavelet transform with corresponding group-velocity curves, the influences of different propagation distances, the features of different source types and the rules of different source depths are examined. It is concluded that the features of AE sources with different propagation distances, types and depths can be obtained by AE technique for rail defect detection. It is very useful to analyze and detect defects in rail defect detection.

Refinement Analysis for Time-Frequency Characteristic of Blasting Seismic Exploration Signals

Accurate extraction of time-frequency features for blasting vibration signals has great significance for blasting seismic exploration, so time-frequency analysis method for blasting seismic signals was researched based on frequency slice wavelet transformation technology, and separation and extraction of time-frequency features were were successfully achieved. Frequency slice wavelet transformation can be introduced into blasting vibration effect analysis fields, it can provide a new research idea for refinement analysis of time-frequency characteristics, and it also has great significance for improving the effect of blasting seismic exploration in China.