Time-frequency Spectrum Research Articles

Multi-sensing signals diagnosis and CNN-based detection of porosity defect during Al alloys laser welding

Porosity is one of the most common defects in laser welding of aluminum (Al) alloys, which seriously affects the welding quality. Online monitoring of porosity defect can provide guidance for the adjustment of process parameters to reduce the occurrence of porosity, which has attracted increasing attention. This paper presents a novel method for recognizing and detecting porosity defect during Al alloys laser welding based on multi-sensing signals diagnosis and deep learning. A multi-sensor platform, including a high-speed camera and a coherent light measurement system, is established to measure the keyhole 3D morphological characteristics. The obtained keyhole depth signal is processed by ensemble empirical mode decomposition (EEMD), and it is found that the change of keyhole depth can be used to recognize the regions where the pores appear. Besides, the sliding window algorithm is used to scan the keyhole opening morphological characteristic signals, followed by the wavelet packet transform (WPT) processing on a small segment of the signals in each window to obtain the time-frequency spectrum graphs. The messy spectrum graphs show that the keyhole opening fluctuates violently, indicating the formation of porosity in the corresponding location of the weld seam. A convolutional neural network (CNN) is constructed to identify the spectrum graphs, realizing the online detection of porosity. The results show that the proposed method has high accuracy and good reliability.

High-Precision and High-Resolution Synchrosqueezing Transform via Time-Frequency Instantaneous Phases

Synchrosqueezing transform (SST) can effectively improve time-frequency precision and resolution by squeezing time-frequency spectra via instantaneous frequencies, and it has been applied in many diverse disciplines; however, the precision of estimated instantaneous frequencies during SST is usually affected by the time-sample interval of the inputted signal; this usually leads to low-precision or inaccurate SST results and limits its further application. To obtain high-precision and high-resolution SST results with high efficiency, we propose a high-precision and high-resolution SST via time-frequency instantaneous phases (HSST); in HSST, time-frequency instantaneous phases with period-jumps removal are used for high-precision instantaneous frequencies estimation and SST. Two synthetic signal examples show that HSST can minimize the impact of the time-sample interval to achieve high-precision and high-resolution SST results with high efficiency. A real 3D seismic data application demonstrates that HSST has fantastic performance in time-frequency precision and resolution enhancement, and it can be widely used in digital signals processing and interpretation fields.

Whirling detection in deep hole drilling process based on multivariate synchrosqueezing transform of orthogonal dual-channel vibration signals

In the deep hole drilling process, whirling vibration is easily initiated due to slender drill bar having larger length to diameter with low torsional and bending stiffness, which may increase the roundness error and roughness of the hole, even damage the tool and the wall of hole. Therefore, whirling detection is an important task to improve part quality and productivity in the deep hole drilling process. Due to the background noise and other disturbances, the whirling feature signal contained in the vibration signal is very puny. The whirling detection in deep hole drilling process based on vibration signal has turn into a challenging task. In this paper, a whirling detection method is proposed based on multivariate synchrosqueezing transform (MSST) of orthogonal dual-channel vibration signals. Firstly, the empirical wavelet transform method is used to extract the high multiple frequency components of the spindle rotation frequency. And then, the MSST is introduced to enhance the whirling feature signal contained in the high multiple frequency components. Finally, singular value decomposition (SVD) method is employed to condense the time–frequency spectrum of MSST and the whirling indicator is constructed based on the singular values. The proposed method is validated with BTA deep hole drilling tests, and the results show the superiority of the proposed method and indicate the proposed method has great potential to be used for the online whirling detection.

Single-Channel FMCW-Radar-Based Multi-Passenger Occupancy Detection Inside Vehicle.

In this paper, we provide the results of multi-passenger occupancy detection inside a vehicle obtained using a single-channel frequency-modulated continuous-wave radar. The physiological characteristics of the radar signal are analyzed in a time-frequency spectrum, and features are proposed based on these characteristics for multi-passenger occupancy detection. After clutter removal is applied, the spectral power and Wiener entropy are proposed as features to quantify physiological movements arising from breathing and heartbeat. Using the average means of both the power and Wiener entropy at seats 1 and 2, the feature distributions are expressed, and classification is performed. The multi-passenger occupancy detection performance is evaluated using linear discriminant analysis and maximum likelihood estimation. The results indicate that the proposed power and Wiener entropy are effective features for multi-passenger occupancy detection.

Assessing the hydroregime of an archetypal riverine wet meadow in the central Great Plains using time‐lapse imagery

AbstractWet meadows are a declining and increasingly degraded ecosystem type. They contribute numerous ecosystem services, including nutrient cycling, water storage, and filtration, and provision of wildlife habitat, particularly for wetland‐dependent species such as the Whooping Crane (Grus americana). Conservation and restoration of wet meadows rely on understanding their hydrology but characterization of wet meadow hydroregimes is difficult given their hydrologic complexity, high variability, and distinct regional differences. To address this challenge, we used ground‐based time‐lapse imagery to assess inundation dynamics of an archetypal wet meadow over a six‐year period in the Central Platte River Valley, Nebraska, USA. We analyzed over 6500 images from March 2011 to May 2017 in the open‐source java‐based image processing software ImageJ. We also obtained data on groundwater, streamflow, precipitation, and evapotranspiration. We assessed the relationship between wet meadow inundation and hydrologic variables using wavelet coherence to look at fluctuations across a time–frequency spectrum and used random forest to identify seasonally specific variables of importance. We found hydroperiod, the duration surface water ponded within the wet meadow, had a mean of 141 d, on average lasting from 10 December to 1 May, but varied annually. Inundation generally peaked in the early spring, on average 10 March, but demonstrated a bimodal distribution, peaking again in late spring during wetter years. While inundation responded rapidly to precipitation events, it was highly related to streamflow, while an elevated groundwater table was necessary for sustained inundation. Overall, our study provided a comprehensive hydrological characterization of a reference wet meadow and demonstrated the utility of time‐lapse cameras for high‐resolution monitoring and assessment of highly variable wetland systems. Considering the uncertainties surrounding land‐ and water‐use changes, climate change, and the increasing demand for freshwater resources by growing human communities, understanding functional wet meadow hydroregimes and interrelated drivers is essential to inform wet meadow restoration, conservation, and management efforts.

Improvement of human error prediction accuracy in single-trial analysis of electroencephalogram.

The prevention of human error is an important task that has already been researched. Previous studies have shown that EEG signals can predict the occurrence of human errors. However, high accuracy has not yet been achieved in a single-trial analysis. This study is aimed to improve the accuracy of single-trial analysis, and propose a method for anomaly detection with auto encoder(AE). In the experiment, we conducted "Press the button(Go)" or "Do nothing(No-Go)" according to the visual stimulus and analyzed the EEG signal from -1000 ms to 0 ms when the stimulus was displayed. We prepared two types of inputs, time series data and frequency spectrum, and an AE was trained to reconstruct the inputs. We then calculated the difference between the reconstructed data and input data and predicted human error by its largeness. In the prediction using Support Vector Machine (SVM) based on the frequency spectrum, some over-fitting occurred and the average accuracy was 43 %. In the prediction using anomaly detection with frequency spectrum was 53 % and could not be classified. The time series data was 63 % which improved the accuracy. A previous study has shown frequency-dependent features such as -band activity and rhythm, as precursors of human error. However, in single-trial analysis, we obtained a higher accuracy by time series data than when by using the frequency spectrum. However, there was no noticeable difference between SVM and anomaly detection methods other than over-fitting. Therefore, in this case, the improvement in accuracy by the anomaly detection method could not be confirmed. However, the result suggests that it is more effective to use the frequency spectrum than the time series data in the single-trial analysis in the future.

Signal Quality Assessment of PPG Signals using STFT Time-Frequency Spectra and Deep Learning Approaches.

Photoplethysmography (PPG) is an important signal which contains much physiological information like heart rate and cardiovascular health etc. However, PPG signals are easily corrupted by motion artifacts and body movements during their recordings, which may lead to poor quality. In order to accurately extract cardiovascular information, it is necessary to ensure high PPG quality in these applications. Although there are several existed methods to get the PPG signal quality, those algorithms are complex and the accuracies are not very high. Thus, this work proposes a deep learning network for the signal quality assessment using the STFT time-frequency spectra. A total of 5804 10s signals are preprocessed and transformed into 2D STFT spectra with 250 × 334 pixels. The STFT figures are as the input of the CNN networks, and the model gives the result as good or bad quality. The model accuracy is 98.3% with 98.9% sensitivity, 96.7% specificity, and 98.8% F1-score. And the heart rate error is much reduced after classification with the reference of ECG signals. Thus, the proposed deep learning approaches can be useful in the classification of good and bad PPG signals. As far as we know, this is the first article using deep learning methods combined with STFT time-frequency spectra to get the signal quality assessment of PPG signals.

Spectral analysis of heart sounds associated with coronary artery disease

Objective. The aim of this study was to find spectral differences of diagnostic interest in heart sound recordings of patients with coronary artery disease (CAD) and healthy subjects. Approach. Heart sound recordings from three studies were pooled, and patients with clear diagnostic outcomes (positive: CAD and negative: Non-CAD) were selected for further analysis. Recordings from 1146 patients (191 CAD and 955 Non-CAD) were analyzed for spectral differences between the two groups using Welch’s spectral density estimate. Frequency spectra were estimated for systole and diastole segments, and time-frequency spectra were estimated for first (S1) and second (S2) heart sound segments. An ANCOVA model with terms for diagnosis, age, gender, and body mass index was used to evaluate statistical significance of the diagnosis term for each time-frequency component. Main results. Diastole and systole segments of CAD patients showed increased energy at frequencies 20–120 Hz; furthermore, this difference was statistically significant for the diastole. CAD patients showed decreased energy for the mid-S1 and mid-S2 segments and conversely increased energy before and after the valve sounds. Both S1 and S2 segments showed regions of statistically significant difference in the time-frequency spectra. Significance. Results from analysis of the diastole support findings of increased low-frequency energy from previous studies. Time-frequency components of S1 and S2 sounds showed that these two segments likely contain heretofore untapped information for risk assessment of CAD using phonocardiography; this should be considered in future works. Further development of features that build on these findings could lead to improved acoustic detection of CAD.

Mortar layer void detection of ballastless track using the impact echo method based on support vector machine

High-speed railway is developing rapidly nowadays. However, under the repeated load of the natural environment and the train, the mortar layer in the ballastless track as the elastic adjustment layer of the structure could be void and cause void defects, thereby seriously affecting the performance of the track board and the safety of the high-speed train. With its simple operation and the use of portable equipment, the impact echo method is a useful nondestructive detection method for detecting defects inside the structure. However, in traditional impact echo detection, manpower is required for judging the final test results, which is highly subjective and time consuming. A field test was designed in this paper, for which a track slab measuring 6.55 × 2.55 × 0.24 m was constructed. Complete cavitation, foam board, and Pykrete were used to simulate the void defects. A 62 × 22 grid with a 0.1 m interval was drawn on the board, each grid point was tapped, and two sensors were placed at the standard position to collect data. Fast Fourier transform was used to obtain the amplitude–frequency domain spectrogram, while wavelet transform was used to obtain the time––frequency spectra. The spectra were processed to extract effective features and form 62 × 22 × 2 = 2,728 sample data. The sample data were divided into a training set and a test set. The training set was used to train the support vector machine (SVM) to classify and identify whether a void defect exists, and the test set was used to evaluate the performance of the SVM. Genetic algorithm (GA) was applied to optimize the parameters in the SVM. Results showed that GA-SVM had a good classification effect with an accuracy of 84.26% and can effectively identify the mortar layer void of the ballastless track slab.

Time-Frequency Spectral Power Assessment of Rolling Element Bearing Faults Using Adaptive Modified Morlet Wavelet Transform

Objectives: The main objectives of this manuscript are to investigate and diagnose rolling element bearing defects in its inception time. Methods: Vibration signal generated by the induction motor contains a series of frequency components that have rich and viable information about bearing health conditions. Recently, the Maximum Energy Concentration (MEC), measure of time-frequency spectrum has been employed to investigate the small variations in low frequency biomedical signal spectrum. In this paper, the above technique has been modified and applied to study the bearing defects of induction motor using vibration signal and it is termed as Adaptive Modified Morlet Wavelet (AMMW) transform. Initially, this proposed method was validated on two medium frequency synthetic time series signals in terms of MEC measurement at different Signal to Noise Ratio (SNR). Results: The simulated results have depicted that the AMMW method provides excellent timefrequency localization capability over other time-frequency methods like Morlet wavelet transform, modified Morlet wavelet transform, adaptive S-transform and adaptive modified S-transform. Then, this method has been applied to the standard database of vibration signal to determine interquartile power for fault detection purpose and also fault index parameter termed as IFI has been analyzed to detect small variation in vibration signals.

Mask filtering to the Wigner-Ville distribution

The Wigner-Ville distribution (WVD) is a high-resolution time-frequency spectral analysis method for nonstationary signals; yet, it suffers from cross-term interference among signal components. We proposed applying a masking filter directly to the WVD time-frequency spectrum to suppress the cross-term interference. Conventional methods for suppressing interference include the smoothed pseudo-WVD (SP-WVD) method, which incorporates smooth filtering in the time and frequency directions. We exploited the SP-WVD spectrum as a reference to design the masking filter; thus, the mask-filtered WVD (MF-WVD) procedure is data-adaptive. The MF-WVD method preserves the high-resolution energy concentration in the spectrum portrayed by the standard WVD, while suppressing the cross-term interference cleanly as in the SP-WVD method. Applying the MF-WVD method to field 3D seismic data generates high-resolution spectral cubes for various frequencies, and these spectral cubes may be used intuitively for detecting reservoir-related characteristics.

Evolutionary Optimization of Machining Parameters Based on Surface Roughness in End Milling of Hot Rolled Steel.

Surface roughness measurements of machined parts are usually performed off-line after the completion of the machining operation. The objective of this work is to develop a surface roughness prediction method based on the processing of vibration signals during steel end milling operation performed on a vertical CNC machining center. The milling cuts were run under varying conditions (such as the spindle speed, feed rate, and depth of cut). This is a first step in the attempt to develop an online milling process monitoring system. The study presented here involves the analysis of vibration signals using statistical time parameters, frequency spectrum, and time-frequency wavelet decomposition. The analysis resulted in the extraction of 245 features that were used in the evolutionary optimization study to determine optimal cutting conditions based on the measured surface roughness of the milled specimen. Three feature selection methods were used to reduce the extracted feature set to smaller subsets, followed by binarization using two binarization methods. Three evolutionary algorithms—a genetic algorithm, particle swarm optimization and two variants, differential evolution and one of its variants, have been used to identify features that relate to the “best” surface finish measurements. These optimal features can then be related to cutting conditions (cutting speed, feed rate, and axial depth of cut). It is shown that the differential evolution and its variant performed better than the particle swarm optimization and its variants, and both differential evolution and particle swarm optimization perform better than the canonical genetic algorithm. Significant differences are found in the feature selection methods too, but no difference in performance was found between the two binarization methods.

Crustal vertical deformation of Amazon Basin derived from GPS and GRACE/GFO data over past two decades

In this study, the Gravity Recovery and Climate Experiment (GRACE) satellite observations, combining 71 continuous Global Positioning System (CGPS) data, are used to detect surface vertical loading deformation of the Amazon Basin during 2002–2020. The results show that the maximal annual amplitude of the surface mass changes derived by GRACE is more than 80 cm in terms of the equivalent water height (EWH) in the Amazon Basin. Most part of Amazon experiences mass gain, especially the Amazon River, while there is little mass loss in the northern and eastern parts. Through the Pearson correlation analysis, the monthly de-trended time series of GPS-observed vertical deformation and GRACE-derived mass loading are in good agreement with an average correlation coefficient of about 0.75 throughout the Amazon region. The common seasonal signals of GPS vertical displacements and GRACE/GFO loading deformations are extracted using the stack averaging. The two kinds of common seasonal signals show a good consistency, and together indicate approximate 20 mm peak-to-peak seasonal amplitude. Strong annual variations are identified both in the monthly GPS and GRACE/GFO data by the wavelet analysis. However, the time-frequency spectrum of GPS has more signal details and more significant semi-annual variations than that of GRACE/GFO. These results may contribute to the understanding of secular crustal vertical deformation in the Amazon Basin.

Contactless tracking of humans using non-contact triboelectric sensing technology: Enabling new assistive applications for the elderly and the visually impaired

Continuous monitoring of human motion and its interaction with the surrounding environment is crucial for healthcare applications such as early diagnosis of neurological disorders, or rehabilitation of vulnerable patients with walking difficulties in hospital and age-care settings. Additionally, it is also necessary to provide assistive care to visually impaired individuals who are prone to falls. Commonly used wearable sensors require on-body components, which may cause discomfort. More importantly, the forgetfulness of people in hospital or aged-care settings or their resistance in using the on-body units, undermine the effectiveness of a wearable-based monitoring strategy. The ideal monitoring system should not interfere with the individual’s privacy and daily routines and work continuously, without requiring any specific on-body devices. Here, we present a contactless sensing platform that can monitor and distinguish complex movements of human subjects, without requiring them to use wearable devices. The approach uses flexible Non-Contact Triboelectric Sensors (NCTS) (7 ×3.6 ×0.5 cm3) produced by attaching a negatively pre-charged PDMS film onto an aluminum (Al) foil. By exploiting the sensitivity to external electrostatic charges, near-field remote monitoring of human movements at distances as far as 1.5 m, is successfully achieved. Furthermore, The NCTS system differentiates activities such as walking, running, and jumping; and estimates walking speed, relative position, and direction of two people walking together in a contactless manner, which is not well examined in the literature of Non-contact Triboelectric Sensors, by analyzing the time domain and frequency spectrum of the output signal. Besides, the sensor is integrated with a portable accident prevention system designed to avoid visually impaired people’s collisions with obstacles. Falling detection tests provide promising results for their application in elderly safety. Finally, indoor navigation/position monitoring and direction recognition are achieved and demonstrated using multiple NCTS, for future potential applications in elder activity tracking and people counting in hospitals.

A machine learning-based underwater noise classification method

We proposed a machine learning-based underwater noise classification method that extracts five underwater noise features (the 1/3 octave noise spectrum level (NL), time–frequency spectrum, power spectral density (PSD), Mel-frequency cepstral coefficient, and Mel filter bank energy) from three domains (the frequency, time–frequency and Mel transform domains). We classified underwater noise using the support vector machine (SVM) and convolutional neural networks (CNN) methods and verified the results using the original data from five classes of typical underwater noise and noise-added data with different signal-to-noise ratios (SNRs). The results show that under different SNR conditions, the classification performance were better with the input features of NL and PSD; when the SNR was −10 dB, the corresponding classification accuracies were 98.95% and 97.65%, respectively. The CNN method outperformed the SVM method for classifying underwater noise, and when the SNR was above −20 dB, the mean classification accuracies of the SVM and CNN methods were 87.8% and 95.6%, respectively.

Power Response and Modal Decay Estimation of Room Reflections from Spherical Microphone Array Measurements Using Eigenbeam Spatial Correlation Model

Modal decays and modal power distribution in acoustic environments are key factors in deciding the perceptual quality and performance accuracy of audio applications. This paper presents the application of the eigenbeam spatial correlation method in estimating the time-frequency-dependent directional reflection powers and modal decay times. The experimental results evaluate the application of the proposed technique for two rooms with distinct environments using their room impulse response (RIR) measurements recorded by a spherical microphone array. The paper discusses the classical concepts behind room mode distribution and the reasons behind their complex behavior in real environments. The time-frequency spectrum of room reflections, the dominant reflection locations, and the directional decay rates emulate a realistic response with respect to the theoretical expectations. The experimental observations prove that our model is a promising tool in characterizing early and late reflections, which will be beneficial in controlling the perceptual factors of room acoustics.

Combination of sparsity of direct waves and detection of signal for distant speech enhancement

A joint dereverberation and denoising framework is proposed for enhancement of distant speech in a conference room. The sparsity of speech multipath propagation is exploited by a microphone array for suppressing late reverberation. The dereverberation speech is transformed to solve a minimization problem where the sparse prediction coefficients are estimated by a least-square iteration algorithm. The residual noise components are removed by iteratively reconstructing the time–frequency spectrum of speech signals. Detection of speech presence is carried out over each cell of the spectrum based on the Bayes rule. The data analysis and the real-time processing results have shown that the proposed joint framework outperforms the existing approaches in terms of objective metrics of speech quality.

Trajectory Analysis of Rising Spheroidal Bubbles in a Rolling Column

The rise of spheroidal bubbles is emulated using a hexapod with periodic angular perturbations to mimic bubble paths in columns onboard seagoing vessels. Bubble trajectories are determined using high-speed camera imaging/ultrasound Doppler velocimetry in vertical, inclined, and rolling columns. The performances of the measurement techniques are compared based on the bubble rise velocity time series and frequency spectra. A hierarchy of reference frames is developed to relate the bubble velocity/acceleration components in the (non-)Galilean frames and to infer from trajectography the lift and drag forces applying to the bubble in the moving-bubble frame. Wall disruption during bubble flights in the rolling column is scrutinized using various geometric/dynamic metrics to underline the dissimilarities between near-wall and centerline bubble path oscillations. Although Coriolis, centrifugal, and Euler accelerations play roles neither on the bubble streamwise rise nor the bubble accelerations, the time-varying character of the buoyancy force brings in dynamic features to the bubble rise in a rolling column that are distinct from those observed in static inclined columns.

Characteristics of audible acoustic signal in the process of laser cleaning of paint on metal surface

Laser cleaning is widely used for paint removal. Audible acoustic waves are generated during the process of laser cleaning of paint. The analytical methods of the time domain, frequency domain and time-frequency domain were adopted in this study. The results show that the time-amplitude (TA) spectra, frequency-amplitude (FA) spectra and time-frequency (TF) spectra have different characteristics at different laser fluences. In the FA spectra, the frequency corresponding to the point of maximum amplitude is 0.04 kHz at the cleaning threshold or above the cleaning threshold. The number of laser pulses that are needed to remove paint with an undamaged substrate can be obtained from the peak numbers of the TF spectra. The TF spectra with the damaged substrate have more peak numbers than the actual number of laser pulses needed to remove paint. By combining the characteristics of the FA spectra and TF spectra, an acoustic monitoring method is established, which can evaluate whether the paint is removed and whether the substrate is damaged. This study provides a new basis for audible acoustic monitoring method in the process of laser cleaning of paint on metal surfaces.

CO2 Leakage Identification Method Based on Complex Time–Frequency Spectrum of Atmospheric CO2 Variation

CO₂ Leakage Identification Method Based on Complex Time–Frequency Spectrum of Atmospheric CO₂ Variation