Search

Smooth variable Structure Filter (SVSF) is a model-based robust nonlinear filtering technique, based on the variable structure concept formulated in a predictor-corrector form. It is used for estimating the states of a system and is robust against noise and modeling uncertainties. It ensures stability in the face of model mismatch resulting from a poor model or fault, at the expense of corrective actions, which cause chattering. The chattering contains mismatch footprints that can be exploited to identify system faults and determine their severity. In this paper, information extraction from chattering is investigated to identify model mismatch based on the spectral contents of the chattering signal. To verify the effectiveness of the developed framework for chattering analysis, two case studies are considered. First, the power spectrum of the chattering signal has been employed to identify mismatch and the potential of recovering the temporal information of the model mismatch from the spectrogram is studied, using Short Time Fourier Transform (STFT) for an underdamped second-order system. Then, the proposed strategy is applied to detect and measure the severity of leakage and friction faults as well as the bulk modulus mismatch in an electro-hydraulic actuator.

Obstructive Sleep Apnea (OSA) is the cessation in breathing that must be identified as early as possible to save the patient’s life. Apart from physical diagnosis, a deep learning model can serve the purpose of detecting the apnea swiftly. The detection largely depends upon biological signals such as ECG, EEG, EMG, etc. Because of the high dimensionality nature of the bio signals, feature extraction is very critical in detecting sleep apnea. Many such feature extraction models were fragile to resolve the complexity issue and failed to reduce the non-robustness nature. To surmount all these issues, a novel adaptive deep learning-based model is designed for detecting the sleep apnea. Here two feature sets have been extracted from the ECG signals: Spectral features through Short Term Fourier Transform (STFT) and QRS analysis followed by an auto encoder to extract the deep temporal features. The novel Artificial Hummingbird Pity Beetle Algorithm (AHPBA) is proposed to choose the optimal features and weight parameters, which assists in concatenation of the two feature sets. Then these fused features were given into Multi Cascaded Atrous based Deep Learning Schemes (MCA-DLS) for classification purpose, then it is further optimized by AHPBA by maximizing the variance. MCA-DLS have performed well compared to classifying the signals individually using One Dimensional Convolutional Neural Networks (1DCNN), Long Short-Term Memory (LSTM) and Deep Neural Networks (DNN) as the average accuracy of MCA-DLS is 94.51% whereas the other three provides an average accuracy of 90.83%, 91.98%, and 93.25% respectively for the considered datasets. By using AHPBA the average accuracy of MCA-DLS was enhanced to 96.4%, which is higher than the conventional optimization techniques which are discussed in the result section.

Background: Diagnosing epileptic seizures using electroencephalogram (EEG) in combination with deep learning computational methods has received much attention in recent years. However, to date, deep learning techniques in seizure detection have not been effectively harnessed due to sub-optimal classifier design and improper representation of the time-domain signal. Methods: In this study, we focused on designing and evaluating deep convolutional neural network-based classifiers for seizure detection. Signal-to-image conversion methods are proposed to convert time-domain EEG signal to a time-frequency represented image to prepare the input data for classification. We proposed and evaluated three classification methods comprising of five classifiers to determine which is more accurate for seizure detection. Accuracy data were then compared to previous studies of the same dataset. Results: We found our proposed model and signal-to-image conversion method outperformed all previous studies in the most cases. The proposed FT-VGG16 classifier achieved the highest average classification accuracy of 99.21%. In addition, the Shapley Additive exPlanations (SHAP) analysis approach was employed to uncover the feature frequencies in the EEG that contribute most to improved classification accuracy. To the best of our knowledge, this is the first study to compute the contribution of frequency components to target seizure classification; thus allowing the identification of distinct seizure-related EEG frequency components compared to normal EEG measures. Conclusion: Thus our developed deep convolutional neural network models are useful to detect seizures and characteristic frequencies using EEG data collected from the patients and this model could be clinically applicable for the automated seizures detection.

Safety, security, and privacy are three critical concerns affiliated with the use of drones in everyday life. Considering their ever-shrinking sizes and capabilities, being aware of drone activities in the vicinity becomes an important surveillance item. Therefore, keeping track of drones and preferably their controllers should be included into the already-existing security measures. In this study, a frequency hopping spread spectrum (FHSS) type drone controller signal detection and emitter direction finding framework is proposed to achieve aforementioned goals. Since drone communications signals generally coexist with other FHSS signals in 2.4 GHz industrial, scientific, and medical (ISM) band, first, a method based on cyclostationarity analysis is proposed to distinguish the drone radio controller signals from other signals utilizing 2.4 GHz ISM band. Then, a variant of short-term Fourier transform is introduced to extract the parameters of detected drone remote controller signals. The correct hopping signals are then aggregated based on the clustered parameters to obtain their combined baseband equivalent signal. Furthermore, the resampling process is applied to reduce the unrelated samples in the spectrum and represent the spectrum with the reconstructed signal, which has a much lower bandwidth than the spread bandwidth. Finally, two different multiple signal classification algorithms are utilized to estimate the direction of the drone controller relative to the receiving system. In order to validate the overall performance of the proposed method, the introduced framework is implemented on hardware platforms and tested under real-world conditions. A uniform linear antenna array is utilized to capture over-the-air signals in hilly terrain suburban environments by considering both line-of-sight and non-line–of-sight cases. Direction estimation performance is presented in a comparative manner and relevant discussions are provided.

We propose an efficient method to estimate various parameters of a helicopter blade including the number Bnum and the length by separating the blade signal using independent component analysis (ICA) and an analytic solution obtained by monostatic/bistatic geometry. The proposed method is composed of two parts, one for even Bnum and one for odd Bnum. For even Bnum the blade parameters are estimated using ICA plus analytic estimation. For odd Bnum the parameters are estimated using two ICAs plus analytic estimation. Experimental results using the radar signal of a blade model calculated by physical optics demonstrate that the proposed method successfully separates the blade signals for both even and odd Bnums and estimates parameters with high accuracy.

In advanced driver assistance systems to conditional automation systems, monitoring of driver state is vital for predicting the driver's capacity to supervise or maneuver the vehicle in cases of unexpected road events and to facilitate better in-car services. The paper presents a technique that exploits millimeter-wave Doppler radar for 3D head tracking. Identifying the bistatic and monostatic geometry for antennas to detect rotational vs. translational movements, the authors propose the biscattering angle for computing a distinctive feature set to isolate dynamic movements via class memberships. Through data reduction and joint time-frequency analysis, movement boundaries are marked for creation of a simplified, uncorrelated, and highly separable feature set. The authors report movement-prediction accuracy of 92%. This non-invasive and simplified head tracking has the potential to enhance monitoring of driver state in autonomous vehicles and aid intelligent car assistants in guaranteeing seamless and safe journeys.

This paper evaluates brain functional connectivity (FC) using Electroencephalography data from 18 healthy subjects using visual oddball paradigm. We have evaluated the effect of different oddball tasks in different time segments. We have also compared the brain FC between the resting state-task data and task-task data. The trials from oddball tasks were averaged to form grand-average event-related potential. Each epoch was divided into six segments with fixed time interval. Coherence was used to compute the pairwise connections between 10-20 electrodes and adjacency matrices were formed. The threshold of 0.9 was set to binarize the matrices and form undirected graphs. Graph parameters (node degree, network density, and betweenness centrality) were used to quantify the networks. Node degree within a brain lobe was summed to form a brain regional activation graph. Node betweenness centrality was used to study the dynamic changes of the hub nodes in different oddball tasks. Edge betweenness centrality was used to quantify the significant functional networks in different oddball tasks. Results show that 1) denser brain network was found in target with response (TR) case as compared with target with no response and no target no response cases especially over 3rd, 4th, and 5th time segments which are before and during P300 period. 2) Higher brain activity in terms of node degree is distributed over all brain regions for TR case than the other two cases. 3) It is also observed that the parietal node is significant for P300 response. 4) Interactions among parietal, central, and temporal regions are significant for motor response that is only observed for TR case. The present findings suggest that our methodology is able to identify differences in FC patterns elicited by different oddball tasks for different brain states in different time segments.

When the power system is suffered from disturbances, such as unbalance, load variation, and fault occurrence, not only the phase and magnitude of the power signal will change but also the fundamental frequency can deviate away from the nominal one up to 5 Hz as per IEEE standard C37.118. As large frequency deviation will lead to great errors on Taylor approximation, the accuracy can be improved by expanding the dynamic characteristics in form of Taylor series near real fundamental frequency of supplied signals instead of nominal frequency (50/60 Hz). The corresponding coefficient against the different frequency estimations is applied to the dynamic phasor estimator to yield accurate synchrophasor estimation with the consideration of large frequency deviation. First, raw phasors attained by short time Fourier transform, are employed to provide the frequency estimation. Then, according to the frequency estimation, the coefficient matrix calculated off-line in form of lookup table is selected to calculate phasor estimation at the center of data window. Finally, a shift process is taken to give phasor estimation at report time. Simulated signals, PSCAD-generated signal, RTDS-generated data, and field data are employed to evaluate the performance of the proposed algorithm. The results show that the proposed algorithm can get more accurate phasor estimations than our previous work at most of the time with the cost of minor increase of computational power. [ABSTRACT FROM PUBLISHER]

In this paper, time–frequency (TF) decomposition (TFD) is studied in the framework of sparse regularization theory. The short-time Fourier transform is first formulated as a convex constrained optimization where a mixed \ell1{-}\ell2 norm of the coefficients is minimized subject to a data fidelity constraint. Such formulation leads to a novel invertible decomposition with adjustable TF resolution. Then, a fast and efficient algorithm based on the alternating split Bregman technique is proposed to carry out the optimization with computational complexity [N^2 \log(N)]. Window length is a key parameter in windowed Fourier transform which affects the TF resolution; a novel method is also presented to determine the optimum window length for a given signal resulting to maximum compactness of energy in the TF domain. Numerical experiments show that the proposed sparsity-based TFD generates high-resolution TF maps for a wide range of signals having simple to complicated patterns in the TF domain. The performance of the proposed algorithm is also shown on real oil industry examples, such as ground roll noise attenuation and direct hydrocarbon detection from seismic data. [ABSTRACT FROM AUTHOR]

In this paper, we propose an alternative efficient method to calculate the Gabor coefficients of a signal given a synthesis window with a support of size much lesser than the length of the signal. The algorithm uses the canonical dual of the window (which does not need to be calculated beforehand) and achieves a computational cost that is linear with the signal length in both analysis and synthesis. This is done by exploiting the block structure of the matrices and using an ad hoc Cholesky decomposition of the Gabor frame matrix. [ABSTRACT FROM AUTHOR]

A fast algorithm for creating time-frequency representations based on a special case of the short-time Fourier transform (STFT) is presented. The algorithm is extended with the method known us time-frequency reassignment. This approach makes time-frequency reassignment well suited for real-time implementations. [ABSTRACT FROM AUTHOR]