Time-frequency Processing Research Articles

A characteristic extraction method for VoicePrint slice statistics base on joint time-frequency processing

Mel-Frequency Cepstral Coefficients (MFCC) and its differential coefficients are widely used as a typical non-linear spectral envelope feature in passive sonar target recognition. MFCC emphasizes the frequency domain information at the low-frequency end of the target, while the differential coefficients add dynamic characteristics of MFCC in the time domain. However, the utilization of time information is not sufficient. In response to this issue, we propose two solutions: one is the characteristic fusion method of Weighted Mel-Frequency Cepstral Coefficient based on Bhattacharyya Distance (BD-WMFCC); the other is the characteristic extraction method of VoicePrint Slice Statistics (VPSS) based on time–frequency joint processing. BD-WMFCC combines weighted differential coefficients with MFCC using Bhattacharyya distance, maximizing the utilization of time information. VPSS is a linear spectral feature that utilizes the stability of the spectral intensity within a certain bandwidth over continuous time. It extracts the steady-state characteristics of the target's spectrum with respect to time changes at equidistant bandwidths. Simulation and real data analysis have shown that VPSS demonstrates outstanding separability and classification performance in both non-linear and linear spectral features, with strong stability. In practical data analysis, the accuracy of VPSS features under the AlexNet algorithm is 3.8 % and 5.7 % higher than the spectral (Fre) and Mel features, respectively.

Measuring compound defect of bearing by wavelet gradient integrated spiking neural network

Compound defect is the common damage of equipment. However, measuring compound defect is difficult because of the complex correlation between multiple single-point defects. Fortunately, as the third generation of neural network that is closest to biological intelligence, the spiking neural network (SNN) has a huge potential on building classifiers. Moreover, time-frequency processing technologies for raw signals are often independent of the training of neural network, so that the end-to-end mode is not able to be built. Therefore, this paper tries to combine the wavelet packet transform (WPT) with SNN, and proposes a wavelet gradient integrated spiking neural network (WGI-SNN) framework. First, a wavelet gradient propagation mechanism (WGPM) is deduced, then recursive down-sampling wavelet networks (DSWNs) is designed to simulate the WPT. Secondly, a SNN model integrating DSWNs is constructed to measure compound defect samples. In addition, an experiment is implemented to verify the effectiveness and practicability of the proposed method.

Experimental study and comparative analysis of pitting fault in spur gear system

This paper uses a dynamic six-degree-of-freedom model that considers torsional and lateral motions to predict the impact of pitting on vibration parameters in a spur gearbox for various operating speeds and torque loads. The study examines the dynamic characteristics of a gearbox with localized pitting damage on a single gear tooth using theoretical and experimental approaches. The research analyzes the forced vibrations of a single-stage spur gear system with pitting damage, which includes variations in mesh stiffness, damping, and gear error excitation, to identify symptoms of default. The equation of motion for the rotary gearbox system is established using the Lagrangian method in tandem with Short-Time Fourier Transform (STFT) and frequency-RPM map fault diagnosis. During real-time vibration monitoring, vibration signals are captured via accelerometers and processed in both the time and frequency domains using the LabVIEW data acquisition signal processing package to extract diagnostic information. The experimental findings demonstrate how vibration analysis combined with time-frequency processing can recognize machine conditions even in harsh operational conditions. Moreover, the experimental results indicate a significant similarity with the theoretical analysis and validate the effectiveness of the RPM frequency technique-based pitting detection method, which can be an asset in gear fault monitoring.

A Multi-Stage Acoustic Echo Cancellation Model Based on Adaptive Filters and Deep Neural Networks

The presence of a large amount of echoes significantly impairs the quality and intelligibility of speech during communication. To address this issue, numerous studies and models have been conducted to cancel echo. In this study, we propose a multi-stage acoustic echo cancellation model that utilizes an adaptive filter and a deep neural network. Our model consists of two parts: the Speex algorithm for canceling linear echo, and the multi-scale time-frequency UNet (MSTFUNet) for further echo cancellation. The Speex algorithm takes the far-end reference speech and the near-end microphone signal as inputs, and outputs the signal after linear echo cancellation. MSTFUNet takes the spectra of the far-end reference speech, the near-end microphone signal, and the output of Speex as inputs, and generates the estimated near-end speech spectrum as output. To enhance the performance of the Speex algorithm, we conduct delay estimation and compensation to the far-end reference speech. For MSTFUNet, we employ multi-scale time-frequency processing to extract information from the input spectrum. Additionally, we incorporate an improved time-frequency self-attention to capture time-frequency information. Furthermore, we introduce channel time-frequency attention to alleviate information loss during downsampling and upsampling. In our experiments, we evaluate the performance of our proposed model on both our test set and the blind test set of the Acoustic Echo Cancellation challenge. Our proposed model exhibits superior performance in terms of acoustic echo cancellation and noise reverberation suppression compared to other models.

Time–frequency analysis of laser speckle contrast for transcranial assessment of cerebral blood flow

Improving noninvasive approaches to monitoring blood microcirculation is an urgent topic in modern biomedical engineering. One of the actively developing methods is laser speckle contrast imaging (LSCI), which allows not only the visualization of microvessels but also the quantitative measurements of microcirculatory blood flow. This work shows the application of LSCI for mapping the cerebral vessels of a laboratory animal, and also presents the time–frequency processing of the registered signal. Thus, we expand the capabilities of the existing LSCI approach and demonstrate spatial mapping of blood flow rhythms. The proposed technology makes it possible not only to measure the relative cerebral blood flow but also to expand diagnostic capabilities for a detailed analysis of the physiological mechanisms of changes in blood flow.

Linear time-invariant (LTI) modeling for aerial and underwater acoustics

Most newcomers to acoustic signal processing understand that linear time-invariant (LTI) filters can remove out-of-band noise from time series signals. What many acoustics researchers may not realize is that LTI models can be applied much more broadly, including to non-linear and time-variant systems. This presentation covers an overview of the autoregressive (AR), moving-average (MA), and autoregressive moving-average (ARMA) family of LTI models and their many useful applications in acoustics. Examples include analytic time-frequency processing of multi-component echolocation signals, fractional-delay filtering for acoustic time series simulations, broadband acoustic array beamforming, adaptive filtering for noise cancelation, and system identification for acoustic equalizers (i.e., flattening the frequency response of a source-receiver pair). This talk serves as a brief tutorial and inspiration for researchers who want to expand their use of signal processing, especially those in the fields of animal bioacoustics, aerial acoustics, and underwater acoustics.

Research on the Classification of ECG and PCG Signals Based on BiLSTM-GoogLeNet-DS

Because a cardiac function signal cannot reflect cardiac health in all directions, we propose a classification method using ECG and PCG signals based on BiLSTM-GoogLeNet-DS. The electrocardiogram (ECG) and phonocardiogram (PCG) signals used as research objects were collected synchronously. Firstly, the obtained ECG and PCG signals were filtered, and then the ECG and PCG signals were fused and classified by using a bi-directional long short-term memory network (BiLSTM). After that, the time-frequency processing was performed on the filtered ECG and PCG signals to obtain the time-frequency diagram of each signal; the one-dimensional signal was changed into a two-dimensional image signal, and the size of each image signal was adjusted to input the improved GoogLeNet network for classification. Then we obtained the two-channel classification results. The three-channel classification results, combined with the classification results of the above BiLSTM network, were finally obtained. The classification results of these three channels were used to make a decision via the fusion strategy of the improved D-S theory. Finally, we obtained the classification results. Taking 70% of the signals in the database as training data and 30% as test data, the obtained classification accuracy was 96.13%, the sensitivity was 98.48%, the specificity was 90.8%, and the F1 score was 97.24%. From the experimental results, the method proposed in this paper obtained high classification accuracy, and the classification effect was better than a cardiac function signal, which makes up for the low accuracy of the cardiac function signal for judging cardiac disease.

Submarine cable fault identification based on FCN-GRU-SVM

With the continuous growth of the global submarine cable laying length, the frequency of faults and the caused economic losses are increasing year by year. Rapid identification of submarine cable faults can reduce economic losses. For several typical submarine cable faults, this paper proposes a recognition method based on FCN-GRU-SVM. First, pre-process the original signal, then import the data into FCN-GRU to extract the signal features, and finally use SVM to identify the fault types. The experimental verification using the data set obtained by finite element simulation shows that the proposed method is superior to the comparison model and other time-frequency processing methods in accuracy and robustness.

Development of an optimization method for measuring the Doppler frequency of a packet taking into account the fluctuations of the initial phases of its radio pulses

The necessity of estimating the decrease in the accuracy of measuring the informative parameters of a radar signal in real conditions of its propagation and reflection has been substantiated. The results of the estimation determine the requirements for optimizing this measurement to achieve the required efficiency. A numerical analysis of the decrease in the accuracy of measuring the Doppler frequency of a coherent packet is presented, depending on the statistical characteristics of fluctuations of the initial phases of its radio pulses. Expressions are given for calculating the fluctuation component of the measurement error of radio pulse packet frequency for various coefficients of interpulse correlation of phase fluctuations. An assessment is made of the possibility of increasing the accuracy of Doppler frequency measurement, which can be ensured by statistical optimization of the algorithm for time-frequency processing of a given radar signal by taking into account its phase fluctuations. The conditions for the multiplicative influence of phase fluctuations of radio pulses of the received packet are substantiated, which determine the efficiency of optimization of Doppler frequency measurement. Based on the results of the study, an optimization method for measuring the Doppler frequency of the packet taking into account fluctuations in the initial phases of its radio pulses is proposed. The accuracy of Doppler frequency measurement under the influence of both the internal noise of the radar receiver and the correlated phase fluctuations of its radio pulses is estimated. The efficiency of optimization of measuring the Doppler frequency of the packet is estimated taking into account fluctuations of the initial phases of its radio pulses by means of computer simulation. It is proved that, under the influence of phase fluctuations, the accuracy of Doppler frequency measurement can be increased due to the performed optimization from 1.86 to 6.29 times. This opens the way to improving the existing algorithms for measuring the higher time range derivatives to improve the quality of tracking complex maneuvering aerodynamic objects. This explains the importance and usefulness of the work for the radar theory.

Respiratory Monitoring Based on Tracheal Sounds: Continuous Time-Frequency Processing of the Phonospirogram Combined with Phonocardiogram-Derived Respiration.

Patients with central respiratory paralysis can benefit from diaphragm pacing to restore respiratory function. However, it would be important to develop a continuous respiratory monitoring method to alert on apnea occurrence, in order to improve the efficiency and safety of the pacing system. In this study, we present a preliminary validation of an acoustic apnea detection method on healthy subjects data. Thirteen healthy participants performed one session of two 2-min recordings, including a voluntary respiratory pause. The recordings were post-processed by combining temporal and frequency detection domains, and a new method was proposed—Phonocardiogram-Derived Respiration (PDR). The detection results were compared to synchronized pneumotachograph, electrocardiogram (ECG), and abdominal strap (plethysmograph) signals. The proposed method reached an apnea detection rate of 92.3%, with 99.36% specificity, 85.27% sensitivity, and 91.49% accuracy. PDR method showed a good correlation of 0.77 with ECG-Derived Respiration (EDR). The comparison of R-R intervals and S-S intervals also indicated a good correlation of 0.89. The performance of this respiratory detection algorithm meets the minimal requirements to make it usable in a real situation. Noises from the participant by speaking or from the environment had little influence on the detection result, as well as body position. The high correlation between PDR and EDR indicates the feasibility of monitoring respiration with PDR.

Deep Learning Assisted Time-Frequency Processing for Speech Enhancement on Drones

This article fills the gap between the growing interest in signal processing based on Deep Neural Networks (DNN) and the new application of enhancing speech captured by microphones on a drone. In this context, the quality of the target sound is degraded significantly by the strong ego-noise from the rotating motors and propellers. We present the first work that integrates single-channel and multi-channel DNN-based approaches for speech enhancement on drones. We employ a DNN to estimate the ideal ratio masks at individual time-frequency bins, which are subsequently used to design three potential speech enhancement systems, namely single-channel ego-noise reduction (DNN-S), multi-channel beamforming (DNN-BF), and multi-channel time-frequency spatial filtering (DNN-TF). The main novelty lies in the proposed DNN-TF algorithm, which infers the noise-dominance probabilities at individual time-frequency bins from the DNN-estimated soft masks, and then incorporates them into a time-frequency spatial filtering framework for ego-noise reduction. By jointly exploiting the direction of arrival of the target sound, the time-frequency sparsity of the acoustic signals (speech and ego-noise) and the time-frequency noise-dominance probability, DNN-TF can suppress the ego-noise effectively in scenarios with very low signal-to-noise ratios (e.g. SNR lower than −15 dB), especially when the direction of the target sound is close to that of a source of the ego-noise. Experiments with real and simulated data show the advantage of DNN-TF over competing methods, including DNN-S, DNN-BF and the state-of-the-art time-frequency spatial filtering.

Remote Compact Seismic Sensor for the Moving Person Detection

The results of the development of a remote compact seismic sensor for the detection of a moving person, as well as sensor testing, are presented in this letter. A choice of a seismic transducer and a seismic signal processing unit is justified. Characteristics of seismic signals from a moving person are analyzed, and an algorithm for the person detection is proposed. As tests have shown, the developed seismic sensor has high-level parameters, which allows it to be successfully used both for reconnaissance purposes and for a various objects guard. It is shown that such parameters can be achieved in a relatively simple device that uses temporal parameters of seismic signals without the use of complex and energy expenditure spectral, time-frequency, statistical, learning, or self-learning information processing methods.

Interference Method for Estimating the Coordinates of a Moving Noise Source in Shallow Water Using High-Frequency Signals

The article presents the results of a high-frequency experiment to localize a moving noise source using a cylindrical small-sized vector–scalar array. Time–frequency processing was used, matched to the interference pattern formed by the source. The time dependences of the bearing, velocity, distance, and depth of the source are reconstructed. The experimental data are qualitatively and quantitative explained using a model two-beam interferogram formed by a direct beam and a beam reflected from the free surface.

Deep Reservoir Computing Meets 5G MIMO-OFDM Systems in Symbol Detection

Conventional reservoir computing (RC) is a shallow recurrent neural network (RNN) with fixed high dimensional hidden dynamics and one trainable output layer. It has the nice feature of requiring limited training which is critical for certain applications where training data is extremely limited and costly to obtain. In this paper, we consider two ways to extend the shallow architecture to deep RC to improve the performance without sacrificing the underlying benefit: (1) Extend the output layer to a three layer structure which promotes a joint time-frequency processing to neuron states; (2) Sequentially stack RCs to form a deep neural network. Using the new structure of the deep RC we redesign the physical layer receiver for multiple-input multiple-output with orthogonal frequency division multiplexing (MIMO-OFDM) signals since MIMO-OFDM is a key enabling technology in the 5th generation (5G) cellular network. The combination of RNN dynamics and the time-frequency structure of MIMO-OFDM signals allows deep RC to handle miscellaneous interference in nonlinear MIMO-OFDM channels to achieve improved performance compared to existing techniques. Meanwhile, rather than deep feedforward neural networks which rely on a massive amount of training, our introduced deep RC framework can provide a decent generalization performance using the same amount of pilots as conventional model-based methods in 5G systems. Numerical experiments show that the deep RC based receiver can offer a faster learning convergence and effectively mitigate unknown non-linear radio frequency (RF) distortion yielding twenty percent gain in terms of bit error rate (BER) over the shallow RC structure.

Emitter signals modulation recognition based on discriminative projection and collaborative representation

To enhance the modulation recognition performance of emitter signals under low signal-to-noise ratio (SNR), a recognition system based on secondary time-frequency distribution, discriminative projection, and collaborative representation is proposed. Firstly, a novel time-frequency processing method, including sparse-domain noise reduction and secondary feature extraction, is proposed to reduce noise interference and information redundancy in time-frequency images. In this way, secondary time-frequency distribution with high stability and detailed representation is obtained. Then, the classifier based on discriminative projection and collaborative representation was designed to enhance the ability of low-dimensional representation and between-class discrimination, which optimised using the mini-batch random gradient descent method. As shown in the simulation, the overall average recognition success rate of this system aiming at eight types of emitter signals reaches 95.6% at the SNR of -8 dB. Results of simulation and analysis indicate the superiority of the proposed classification system in terms of robustness, timeliness, and adaptability.

EVALUATION OF THE POSSIBILITIES OF PROVIDING THE NECESSARY ACCURACY OF THE RADIAL VELOCITY MEASURING OF A TARGET BY COHERENT-PULSE RADARS OUTSIDE THE LINE OF SIGHT

The purpose of the article. The article estimates the possible values of the components of the mean-square error of measuring the radial velocity of the target which appear as a result of fluctuations of the phase of the radio signal outside the radar line of sight. The expediency of using a coherent burst of radio pulses to provide the necessary detection range with specified quality indicators is substantiated. Consideration is carried out for a model of a signal with a random amplitude and an initial phase. It is assumed that phase fluctuations are distributed according to the normal law with zero mean, and their correlation decreases with an increase in the interval between the radio bursts of the packet alternating. Results. The results show that due to the phase fluctuations of the radio pulses of the received pack, the mean-square measuring error of the target radial velocity can exceed the values determined by the tactical requirements for coherent-pulse radars. Conclusions. The performed numerical analysis allows to determine the degree of reduction of the quality of the radio pulses burst time-frequency processing in coherent-pulse radars and evaluates the degree of reduction of the effectiveness of the further secondary processing of the radar information.

Target detection in pseudo Wigner-Ville distribution of underwater beamformed signals

Because of clutter or reverberation in the water, it is difficult to distinguish a target from them with a beamforming image. It is well-known that the pseudo Wigner-Ville distribution, which is high-resolution time-frequency processing, can detect target signals effectively. In this study, the pseudo Wigner-Ville distribution with the beamformed signals was analyzed to detect an underwater target. The acoustic signals were measured using an array system consisting of a transmitting transducer and a line array of 16 elements at sea. The target was an air-filled aluminum spherical shell with a thin thickness. The transmitted signal was 3 to 8 kHz linear frequency modulation continuous waveforms. The experiment results show that the target can be easily identified with the pseudo Wigner-Ville distribution. [Work financially supported by the research project PES3180 funded by KRISO.]

A Comparative Analysis of Artificial Intelligence-Based Methods for Fault Diagnosis of Mechanical Systems

Abstract The present research studied fault diagnosis of composite sheets using vibration signal processing and artificial intelligence (AI)-based methods. To this end, vibration signals were collected from sound and faulty composite plates. Using different time-frequency signal analysis and processing methods, a number of features were extracted from these signals and the most effective features containing further information on these composite plates were provided as input to different classification systems. The output of these classification systems reveals the faults in composite plates. The different types of classification systems used in this research were the support vector machine (SVM), adaptive neuro-fuzzy inference system (ANFIS), k-nearest neighbor (k-NN), artificial neural networks (ANNs), Extended Classifier System (XCS) algorithm, and the proposed improved XCS algorithm. The research results were reflective of the superiority of ANFIS in terms of precision, while this method had the highest process duration with an equal number of iterations. The precision of the proposed improved XCS method was lower than that of ANFIS, but the duration of the process was shorter than the ANFIS method with an equal number of iterations.

Unified cochlear model for low- and high-frequency mammalian hearing

The spatial variations of the intricate cytoarchitecture, fluid scalae, and mechano-electric transduction in the mammalian cochlea have long been postulated to provide the organ with the ability to perform a real-time, time-frequency processing of sound. However, the precise manner by which this tripartite coupling enables the exquisite cochlear filtering has yet to be articulated in a base-to-apex mathematical model. Moreover, while sound-evoked tuning curves derived from mechanical gains are excellent surrogates for auditory nerve fiber thresholds at the base of the cochlea, this correlation fails at the apex. The key factors influencing the divergence of both mechanical and neural tuning at the apex, as well as the spatial variation of mechanical tuning, are incompletely understood. We develop a model that shows that the mechanical effects arising from the combination of the taper of the cochlear scalae and the spatial variation of the cytoarchitecture of the cochlea provide robust mechanisms that modulate the outer hair cell-mediated active response and provide the basis for the transition of the mechanical gain spectra along the cochlear spiral. Further, the model predicts that the neural tuning at the base is primarily governed by the mechanical filtering of the cochlear partition. At the apex, microscale fluid dynamics and nanoscale channel dynamics must also be invoked to describe the threshold neural tuning for low frequencies. Overall, the model delineates a physiological basis for the difference between basal and apical gain seen in experiments and provides a coherent description of high- and low-frequency cochlear tuning.

S-RVoG Model Inversion Based on Time-Frequency Optimization for P-Band Polarimetric SAR Interferometry

This paper investigates the potential of the time-frequency optimization on the basis of the sublook decomposition for forest height estimation. The optimization is deemed to be capable of extracting a relatively accurate volume contribution when P-band polarimetric interferometric synthetic aperture radar (Pol-InSAR) systems are adopted to observe forest-covered areas. The highest and the lowest phase centers acquired by the time-frequency optimization modify the conventional three-stage inversion process. This paper presents, for the first time, a performance assessment of the time-frequency optimization on P-band Pol-InSAR data over boreal forests. Simultaneously, to alleviate the model inversion errors caused by topographic fluctuations, forest height is estimated based on the sloped Random Volume over Ground (S-RVoG) model in which the incidence angle is corrected with the terrain slope. The E-SAR P-band Pol-InSAR data acquired during the BIOSAR 2008 campaign in Northern Sweden is utilized to evaluate the performance of the proposed method. From the results of the forest height estimation preprocessed with time-frequency optimization, the root mean square error (RMSE) of Random Volume over Ground (RVoG) and S-RVoG model on negative slope are 5.09 m and 4.71 m, respectively. It is concluded that the time-frequency processing and negative terrain slope compensation improve the inversion performance by 41 . 49 % and 11 . 96 % , respectively.