Articles in press have been peer-reviewed and accepted, which are not yet assigned to volumes /issues, but are citable by Digital Object Identifier (DOI).
Display Method:
Vol. 48 Issue 5 2023 Catalog Cover
Display Method:
Frequency-shift compensation localization method for large aperture horizontal array in shallow water
MEI Xiaohan, ZHANG Bo, PENG Zhaohui, YU Xiaotao, ZHAI Duo
2023, 48(5): 901-910.   doi: 10.12395/0371-0025.2022032
Abstract FullText HTML PDF

The interference of different normal mode makes the horizontal longitudinal correlation of sound field oscillate. The frequency-shift compensation relationship between the received signals at different positions is derived by using the horizontal distance-frequency interference structure, combined with the characteristics of high dispersion of spatial distribution of large aperture array elements, a localization method suitable for slowly varying spectrum signal is proposed. By using the change of the frequency-shift compensation with the position of sound source, the ambiguity plane of two elements is superimposed to realize the horizontal two-dimensional plane localization. The simulation results show that the method has good localization performance and good tolerance to environmental parameter mismatch. The linear phase relationship after frequency-shift compensation effectively improves the correlation between the received signals, thereby improving the processing gain of the large aperture array. The aperture advantage of the array improves the spatial resolution, and the peak-to-background ratio of the ambiguity plane is high. Sea trial data validation shows that the average relative deviation of 10−80 km ranging results is 5.68% and the average distance deviation of localization results in two-dimensional plane is 0.78 km.

Array aperture extension method using covariance matrix fitting
LIU Kaixin, FU Jin, ZOU Nan, ZHANG Guangpu, HAO Yu
2023, 48(5): 911-919.   doi: 10.12395/0371-0025.2022074
Abstract FullText HTML PDF

An array aperture extension method based on covariance matrix fitting is proposed to improve the resolution of uniform linear array with small aperture. The relationship between the covariance matrices of signals received by arrays with different number of array elements is analyzed. According to this relationship, an array extension optimization algorithm is constructed to fit the covariance matrix of a large aperture virtual array by using the covariance matrix of a small aperture actual array. Simulation and lake experiment results show that using the fitted covariance matrix in existing DOA estimation methods can reduce the beamwidth and improve the resolution. As the number of virtual array elements increases, the resolution synchronously improves. This method can be used to improve the DOA estimation performance when the array aperture is insufficient or the signal-to-noise ratio is reduced.

Signal space transform and multidimensional information joint processing for moving target
SHI Yunjia, PIAO Shengchun, GUO Junyuan, ZHANG Liu
2023, 48(5): 920-936.   doi: 10.12395/0371-0025.2022015
Abstract FullText HTML PDF

A signal space transform method for detection of line spectrum signals radiated by moving target is proposed. With time-frequency analysis and beamforming processing, a three-dimensional (3D) time-frequency-azimuth signal space is obtained in the array coordinate system. Given certain target course angle, this 3D signal space is transformed to the time-frequency-relative bearing spectrum signal space in the target coordinate system. The findings demonstrate that the method is capable of converting the target line spectrum signal 3D trajectory to a 2D trajectory on the corresponding slice in the transformed signal space, and signal processing gain of multi-dimensional information joint processing is obtained easier than in 3D signal space.

Direction of arrival estimation of acoustic vector sensor array based on the combined information processing of pressure and particle velocity using covariance matrix decomposition
ZHANG Xu, SHI Shengguo, ZHU Xiaochun, XU Fujia
2023, 48(5): 937-949.   doi: 10.12395/0371-0025.2022005
Abstract FullText HTML PDF

A combined processing method of pressure and particle velocity (PV-CPM) for acoustic vector sensor array using covariance matrix decomposition (CMD) is proposed. In this method, the cross-covariance matrix of pressure and particle velocity is decomposed into an observation angle coefficient matrix and a residual covariance matrix. To avoid choosing the observation angle, the coefficient matrix is combined with the guidance vector. By adopting the singular value decomposition, the residual covariance matrix is reconstructed into the new Hermitian covariance matrix, which is ultimately used to implement the MVDR beamforming method. Theoretical analysis shows that the new covariance matrix can achieve higher array processing gain. Furthermore, the numerical simulation proves that the computational complexity of the proposed method is similar to that of the Nehorai’s method, but its array processing gain and multi-target resolution are higher than those of the traditional PV-CPM.

Active recognition of underwater targets using GAF-D3Net deep learning network
ZHOU Tian, SI Jikun, DU Weidong, YU Xiaoyang
2023, 48(5): 950-958.   doi: 10.12395/0371-0025.2022016
Abstract FullText HTML PDF

A method for active recognition of underwater targets based on Gramian angular field (GAF) and convolutional neural network (CNN) is proposed. GAF is used to encode the target echo signal into a two-dimensional image, and a lightweight convolutional neural network GAF-D3Net are bulit by dilated convolutions to achieve feature extraction and classification recognition of the target. Experiment results show that the classification accuracy of the proposed method is significantly improved to 99.65% compared to traditional image feature-based classification methods. In the generalization test, comparing the classical CNN migration learning method using sonar images, the area under curve (AUC) of the proposed method reaches 89%, verifying that the proposed method has better generalization performance as well as anti-interference capability. It provides a reliable method for achieving active recognition of underwater targets.

Measurement-driven adaptive likelihood passive weak target tracking
QI Bin, TIAN Jin, ZOU Nan, LIANG Guolong
2023, 48(5): 959-970.   doi: 10.12395/0371-0025.2022029
Abstract FullText HTML PDF

Based on the theory of random finite sets, the track-before-detect for passive sonar is studied, and the adaptive weighted spatial spectrum of multiple signal classification (MUSIC) method is used as pseudo-likelihood ratio function to study the MUSIC-based approximate multi-Bernoulli filtering algorithm. Aiming at the problem of slow response speed of the algorithm to target regeneration, a measurement-driven target regeneration model is proposed. The simulation results show that the proposed algorithm has better tracking performance and less computation than the traditional algorithm at low SNR, and the improved model can significantly reduce the response time of the algorithm to the new target, which is improved by more than 50%. Experimental results show that the proposed method has strong robustness and can track multiple targets accurately at low SNR.

Radiated noise measurement method of underwater low noise target by a vertical nested array
LI Xuegang, JIANG Guoqing, SUN Guocang, HUA Runan, ZHU Jie
2023, 48(5): 971-977.   doi: 10.12395/0371-0025.2022187
Abstract FullText HTML PDF

A radiated noise measurement method of underwater acoustic targets with relatively low power by a vertical nested array is studied. Firstly, the array gain requirements for different frequency bands are obtained through comparative analysis of the background noise level and the radiated noise level of the target. Secondly, an ultra wideband constant-beamwidth beamforming method with high robustness is established for the waveguide environment based on convex optimization. The radiated noise measurement and the following spatial array signal processing procedure is formulated. Finally, a vertical nested array measurement system is developed, and the radiated noise measurement method is verified using lake test data. Experimental results show that the radiated noise measurement method has large working bandwidth, high robustness and low measurement uncertainty of about 2 dB.

Aerodynamic sound radiation of low Mach number flows around a cylinder and truncated cylinders
CAI Jiancheng, YU Mengyao, CHEN Chaoqian, XIA Wenjun, E Shiju
2023, 48(5): 978-988.   doi: 10.12395/0371-0025.2022049
Abstract FullText HTML PDF

Taking the sound radiation of flow past cylinders as the research objective, the quantitative prediction methods of aerodynamic sound in low Mach number flows are discussed. Firstly, the intrinsic relationship between the Lighthill’s acoustic analogy and an acoustic perturbation equation based on the acoustic/viscous splitting technique, is analyzed. Secondly, based on computational fluid dynamics, the two-dimensional (2D) simulations of the subcritical flow around a cylinder and two truncated cylinders are carried out, and performances of the two turbulence models are discussed. Finally, the aerodynamic sound radiations are predicted using the FW-H equation in the Lighthill’s acoustic analogy theory as well as the non-homogeneous wave equation of the acoustic perturbation theory, and the acoustic directivity and spectral characteristics are analyzed. The numerical results show that the acoustic radiations predicted by the FW-H equation and the acoustic perturbation equation are similar in the acoustic spectra with the larger amplitude by the FW-H equation. The truncation of the cylinder at the rear part makes the wake flow fluctuations increase resulting in more intensive sound radiation.

Low-frequency hydrodynamic noise calculation of underwater vehicle with sparse grid conditions
SHEN Qi, LIU Jin, GAO Yan, TAN Zhiqiang, YU Mengsa
2023, 48(5): 989-995.   doi: 10.12395/0371-0025.2022055
Abstract FullText HTML PDF

A computational model of the acoustic vibration response of complex structures excited by pressure fluctuations in the turbulent boundary layer is constructed by combining the structural finite element method, the acoustic finite element method, and non-reflecting boundary conditions. The issue associated with the mesh size of the finite element model and the spatial correlation scale of the pressure fluctuations in the turbulent boundary layer is solved by introducing the virtual grid refinement to accomplish the correction of the uncorrelated flow excitation loads. The underwater vehicle’s low-frequency hydrodynamic noise calculation is carried out under sparse grid settings that meet the demands of the acoustic vibration response. The typical underwater rotary vehicle model is used as an example to calculate its low-frequency hydrodynamic radiation noise. When compared to the results of the discrete subunit method calculation and the large cavitation channel test, the three results are generally consistent, with the deviation of the results in the overlapping frequency range being less than 3 dB. The method is highly applicable and takes into account calculating efficiency and accuracy.

A hybrid method for ultrasonic echo of defects using spatial impulse response and finite element method
DONG Ming, YU Xiangjun, MA Hongwei, CHEN Yuan, CAO Xiangang, WAN Xiang, WANG Xin
2023, 48(5): 996-1003.   doi: 10.12395/0371-0025.2022037
Abstract FullText HTML PDF

The numerical simulation modelling is challenging due to the increased mesh requirements to resolve the large propagation distance. A hybrid modeling method for ultrasonic echo of defects using spatial impulse response and finite element method is proposed. The local area around defect is meshed and the interaction between the ultrasonic wave and the defect is calculated by the finite element method. The vibration velocities of the nodes on the boundary are calculated by spatial impulse response method, which is considered as excitation of the finite element model. The velocity potential of the reflected wave at the receive transducer is also obtained by spatial impulse response method. This method enables efficient prediction of the response of complex scatterers. A simulation example of flat bottom hole is established by the hybrid method and the finite element method respectively, and both the waveforms are consistent. The total time for running the hybrid method is 5.9% against the finite element method, and the runtime for the finite element simulation can be reduced significantly. The echoes of cracks with different depth at shoulder root of idler shaft are simulated by the hybrid method. Echoes reflected from the shoulder and crack are overlapped, meaning crack information is lost. The results show the echo amplitude increases with the crack depth, and the echo amplitude can be used to size the root crack. The hybrid method proposed in this paper lays a foundation for improving the computational efficiency of large-scale finite element models.

Real-time monitoring of acoustic cavitation in high intensity focused ultrasound based on phase characteristics of transducer electrical signals
LIU Chuang, LIU Li, ZHU Qiaomiao, LI Yanhao
2023, 48(5): 1004-1011.   doi: 10.12395/0371-0025.2022113
Abstract FullText HTML PDF

Acoustic cavitation induced by high intensity focused ultrasound (HIFU) can accelerate the thermal ablation of the target tissue. However, real-time monitoring of acoustic cavitation caused by HIFU is an urgent problem to be solved. The phase characteristics of the electrical signal of the HIFU transducer are established to analyze the real-time monitoring of acoustic cavitation. Under different HIFU excitation voltages, experimental research about real-time monitoring of acoustic cavitation in isolated bovine heart tissue irradiated by HIFU has been carried out. In addition, the phase difference of driving electrical signal is compared with the grayscale change of the B-ultrasound image and change results of subharmonic and broadband noise detected by broadband hydrophone. The research results show that when acoustic cavitation occurs, the change of phase difference of driving electrical signal has good consistency with the change of subharmonic and broadband noise detected by hydrophone. By the change of phase difference, real-time and accurate monitoring of the duration of acoustic cavitation occurring in target tissue irradiated by HIFU can be achieved, which provides a promising solution for real-time monitoring of acoustic cavitation caused by HIFU.

Measurements of normal incidence sound absorption coefficient for non-standard sized samples in impedance tubes
LOU Huading, MIN Hequn
2023, 48(5): 1012-1020.   doi: 10.12395/0371-0025.2022025
Abstract FullText HTML PDF

A measurement method is proposed for the normal incidence sound absorption coefficient of non-standard sized samples in impedance tubes, and the corresponding parameters are systematically addressed as well. At first, a type of parallel-arranged acoustic material (PAM) is applied to fill the non-standard sized sample to be a flat inhomogeneous acoustic impedance sample (IAIS) with the standard size in impedance tubes. Secondly, the surface acoustic impedance of IAIS is obtained in impedance tubes according to GB/T 18696.2—2002, and then the surface acoustic impedance and absorption coefficient of the non-standard sample is evaluated with the proposed method based on the equivalent circuit method. A parameter study is conducted in numerical simulations and experimental measurements. Results show that, the accuracy of the proposed method improves with the increase of the area ratio of the sample to impedance tube cross section, and improves with the decrease of the length of the scattering edge. It is also found that, for porous material samples, a non-rigid PAM with surface impedance close to that of the sample is beneficial to the measurement accuracy; while for resonant absorbing samples, the measurement accuracy can stay reasonable even if a rigid PAM is applied.

Sound insulation measurement of building components combining diffuse acoustic field excitation and near-field acoustic holography reconstruction
WANG Hongwei, XIONG Wei, WANG Qiaochu, ZHANG Guangyao, YANG Chenxi
2023, 48(5): 1021-1035.   doi: 10.12395/0371-0025.2022027
Abstract FullText HTML PDF

A method for measuring the airborne sound insulation of building components combining diffuse acoustic field (DAF) excitation and near-field acoustic holography (NAH) radiation intensity reconstruction is proposed. In this method, DAF is first used to excite the vibration of the component and obtain the incident sound power, then the normal sound intensity distribution with high spatial resolution on the surface of the component is reconstructed from the radiation sound pressure field using NAH, and finally, according to the sound intensity distribution, the radiated sound power is calculated and the radiation heat areas are located, so as to realize the measurement of sound insulation and sound insulation defects of components. The experimental research in a sound insulation room shows that under the condition that the test distance and sampling interval are both 0.04 m, the error of the sound insulation measured by this method compared to the sound pressure method is less than 3.3 dB in the 100–5000 Hz frequency band and less than 1.3 dB in the 250–3150 Hz frequency band, and the positioning accuracy of round holes (diameter: 8 mm) and rectangular slots (length: 80 mm, width: 3 mm) is up to the centimeter level. The method has strong stability under the influence of certain amount of reverberation and background noise. When reverberation time in the receiving room increases from 1.0 s to 3.4 s (step size: 0.6 s), and the signal-to-noise ratio decreases from 10 dB to 0 dB (step size: 5 dB), the sound volume measurement errors are within 0.8 dB and 0.3 dB, and the defect location errors are within 0.037 m and 0.035 m. The proposed method helps to improve the measurement capability of sound insulation characteristics of building components in the laboratory, and it is robust to the receiving room test environment.

Delay estimation using encoder-temporal modeling structure for acoustic echo cancellation
LIU Yang, YANG Feiran, YANG Jun
2023, 48(5): 1036-1044.   doi: 10.12395/0371-0025.2022045
Abstract FullText HTML PDF

A delay estimation method based on encoder-temporal modeling structure is proposed to estimate the delay of microphone signal relative to the far-end signal in acoustic echo cancellation. In the proposed method, the far-end signal and the microphone signal in the short-time Fourier transform domain are used as input features. High-dimensional features with phase information are extracted by an encoder composed of complex convolutional neural networks. The memory ability of recurrent neural network is used to learn the time delay relationship between two input signals. A mapping from signal to delay is constructed by the proposed method. The simulation results show that the proposed method has the following advantages over WebRTC-DE and GCC-PHAT: (1) the number of parameters and computational complexity of the model are not affected by the delay; (2) the convergence time and tracking time of delay estimation are effectively reduced; (3) smaller and more stable estimation error and standard deviation are achieved in the case of long reverberation time and double-talk. Experiments on adaptive echo cancellation cascaded with the proposed delay estimation module verify the effectiveness of the new method.

Localization of multiple speakers in the spherical harmonic domain by robust order-aware pseudo-intensity vectors exploiting time-frequency correlation
GAO Weixia, CHEN Huawei
2023, 48(5): 1045-1059.   doi: 10.12395/0371-0025.2022085
Abstract FullText HTML PDF

Multiple sound source localization using pseudo-intensity vector is known sensitive to noise and room reverberation. To deal with the problem, a robust sound localization approach is proposed using pseudo-intensity vector via constructing an order-aware factor of spherical harmonics (OAFSH) by employing signal correlation property in the time-frequency domain. Unlike its existing counterpart, the proposed OAFSH fully takes advantage of the correlation of the eigenbeams that belong to the same source between adjacent time-frequency bins. Theoretical analysis shows that the proposed OAFSH performs better than the existing counterpart in suppressing noise and room reverberation. Simulation results demonstrate that the sound source localization accuracy of the proposed approach is 1.3° ~ 1.9° higher than that of the existing approach under the condition of a 10 dB signal to noise ratio and reverberation time of 0.4 ~ 1.0 s. Moreover, the computational complexity of the proposed approach is also reduced by 25%. Finally, the effectiveness of the proposed approach is further verified by the real-world experiments in practical room environment.

Light-weight speech separation based on dual-path attention and recurrent neural network
YANG Yi, HU Qi, ZHANG Pengyuan
2023, 48(5): 1060-1069.   doi: 10.12395/0371-0025.2022044
Abstract FullText HTML PDF

A light-weight speech separation algorithm based on dual-path attention and recurrent neural network is proposed. First, optional branch structures based on dual-path attention mechanism and dual-path recurrent network are utilized to model the speech signals, which facilitate the extraction of deep feature information and the reduction of training parameters. Second, sub-band processing approach is introduced to alleviate the computation burden. As shown by the experimental results on the LibriCSS dataset, the average word error rate obtained by the proposed algorithm is 8.6% with only 0.15 MiB training parameters and 15.2 G/6s computation cost, which is 3.3−391.3 and 1.1−3.2 times smaller than other mainstream approaches. This proves the proposed algorithm can effectively reduce the training parameters and computation cost while achieving high speech separation performance.

Noise robust voice conversion with the fusion of Mel-spectrum enhancement and feature disentanglement
CHEN Lele, ZHANG Xiongwei, SUN Meng, ZHANG Xingyu
2023, 48(5): 1070-1080.   doi: 10.12395/0371-0025.2022093
Abstract FullText HTML PDF

A novel noise-robust voice conversion model named MENR-VC which combines Mel-spectrum enhancement and feature decoupling is proposed in this paper. Speech content, fundamental frequency, and speaker identity vector features are extracted by three encoders in the model, and mutual information is introduced as a correlation metric to achieve speaker identity conversion via minimizing mutual information for feature decoupling. To overcome the limitations of noisy speech, a deep complex recurrent convolutional network is employed by the model to enhance the noisy Mel spectrum, which serves as input to the speaker encoder. Additionally, the Mel-spectrum enhancement loss function is introduced during the training process to improve the overall loss function of the model. The simulation results demonstrate that similar optimal noise-robust voice conversion methods are outperformed by the proposed model, with an enhancement of 0.12 and 0.07 in the average opinion scores of speech naturalness and speaker similarity of the converted speech respectively. The training of deep neural network can be easily converged when noisy speech is used as training data in the proposed voice conversion model, and the quality of the converted speech is also satisfatory.