To enable precise target positioning in constrained underwater platforms with limited storage,we propose a 1-bit quantized frequency diverse array multiple-input multiple-output(FDA-MIMO)sonar system.The FDA-MIMO sonar sig-nificantly enhances target localization by providing both range and angular resolution.However,the coupling between range and angle in traditional FDA-MIMO processing leads to parameter estimation ambiguity,and grid-based compressed sensing(CS)methods suffer from grid mismatch,resulting in reduced accuracy.To overcome these challenges,we present a 1-bit gridless CS-based target positioning algorithm for accurate underwater localization.First,we develop an array expansion model to decouple the range and angle parameters.Then,leveraging gridless CS with 1-bit quantization,we reformulate the joint estimation as an atomic norm minimization problem.Using convex relaxation,we convert this into an asymmetric cone programming problem and propose an efficient,fast interior-point method for rapid optimization.Numerical simulations show that the proposed algorithm excels in both estimation accuracy and computational speed.
The difference in tonal amplitude fluctuations between surface and submerged sources in shallow ocean can be used for binary source depth discrimination.However,under low signal-to-noise ratio(SNR)conditions,such fluctuation-based discrimination is often compromised by strong background noise.To improve the depth discrimination capability for weak tonal sources,this paper proposes a source depth discrimination method with a passive horizontal line array(HLA)using depth-modulated amplitude fluctuation.We investigate the physical mechanism of depth-modulated amplitude fluctuation in the beam domain and reveal that existing HLA-based decision metrics are sensitive to the amplitude fluctuation induced by modal interferences.To overcome this limitation,a source range-independent metric:the amplitude fluctuation energy ratio(AFER)via the Fourier transform,is proposed.The AFER leverages the distinct frequency characteristics of modal interference fluctuation,depth-modulated fluctuation,and noise fluctuation in the time domain.The combined array gain and integration gain of the AFER are both utilized to suppress noise.Simulation results demonstrate AFER's advantages in weak tonal source(SNR at hydrophone equals-5 dB in 1 Hz band)discrimination under unknown source range conditions compared to the existing decision metric built by the ratio of advanced WISPR(Wagstaff's integration silencing processor)summation(AW-SUM).The analysis of SWellEx-96 experimental data further validates the effectiveness of the proposed method.
Ranging underwater acoustic sources in a waveguide perturbed by internal solitary waves is a challenging problem.This paper investigates the effects of internal solitary waves on the waveguide invariant.Utilizing the properties of the adiabatic component of the acoustic intensity when the internal solitary waves are in motion,a multi-frame finite-rate-of-innovation algorithm is proposed to estimate the original two-dimensional frequencies and waveguide invariant simultaneously from the corrupted acoustic interference pattern.A technique for pairing the two-dimensional frequencies is developed to estimate the range of the source.Acoustic simulations validate the accuracy and robustness of the ranging method in a waveguide with internal solitary waves,and the proportion of estimation error below 0.9 km is almost 80%.
To mitigate the masking of weak target echoes by cumulative reverberation in continuous active sonar(CAS),and to achieve real-time reverberation suppression under the constraints of short detection windows,this paper proposes a se-quential subband sparse filtering with QR decomposition(QR)updates method based on low-rank and sparse decomposition.By leveraging the inherent sequential nature of CAS subband outputs,a low-rank background model is constructed via incremental QR updates.This framework allows for the rapid extraction of sparse components from each subband,effectively separating steady-state reverberation,reverberation fluctuations,and target echoes.Validation using shallow-water experi-mental data demonstrates that the proposed method achieves rapid convergence with only a few subband inputs.It significantly reduces interference from reverberation fluctuations,maintains the structural continuity of targets under low signal-to-noise ratio conditions,and improves the detectability of weak echoes.The online separation framework ensures superior compu-tational efficiency while enhancing system adaptability to complex reverberant environments.
Interference-structure-based passive source localization provides computational efficiency and physical interpret-ability but suffers performance degradation for weak acoustic targets,where low signal-to-noise ratios(SNRs)obscure the observable interference patterns necessary for reliable depth estimation.Narrowband time-window limitations and environ-mental mismatches further introduce blind zones and nonlinear distortions between the source depth and the measured interference structure.To mitigate these effects,this study proposes a hybrid depth estimation method that integrates arrival-angle interference features with a deep learning(DL)framework based on a residual network.Sound intensity in the beam domain is employed to exploit multidimensional and nonlinear relationships in the acoustic field,thereby enhancing robustness under weak target and low SNR conditions.Simulation and sea experiment results demonstrate that the proposed method achieves improved depth estimation accuracy compared with conventional interference-based techniques.The findings indicate that the hybrid interference-DL approach effectively extends the applicability of interference-structure-based localization to weak acoustic targets.
To address the challenge of extracting weak line spectrum from spectrograms,this paper proposes a line spectrum extraction network(LSE-Net)with a simple structure and excellent performance.This network is built on a classic encoder-decoder architecture.To improve training efficiency with a small number of samples,this paper embeds adapters into each layer of the encoder to achieve parameter-efficient fine-tunin,enabling the model to complete training with only a small amount of data and attain excellent performance.To strengthen the line spectrum feature extraction capability,a multi-receptive field feature fusion module is integrated at the skip connections to capture line spectrum information at different scales;meanwhile,a reverse cross attention module is introduced in deep feature processing to accurately enhance the edge features of line spectra and effectively suppress background noise.Experimental results demonstrate that the proposed LSE-Net outperforms state-of-the-art methods(e.g.,Deep Lofargram,Trans-UNet)in low signal-to-noise ratio(SNR)scenarios(ranging from-16 dB to-22 dB).Specifically,at-22 dB SNR,LSE-Net achieves a probability of detection of 0.965,a false alarm rate of 0.373,and a line localization accuracy of 0.786,which are 57.1%,42.3%,and 85.8%higher than those of Deep Lofargram,respectively.The proposed network provides an efficient and reliable solution for weak line spectrum extraction in spectrograms.
High-resolution parameter estimation of underwater targets is a key technology for an active sonar.However,for wideband signals adopted by the active sonar,the Doppler frequency shift associated with the varying Doppler scale factor will cause echo spreading and distortion.This will inevitably degrade the parameter estimation performance when handling the time-domain overlapped echoes from multiple highlights.To solve the problem,this paper proposes a high-resolution parameter estimation based on matrix mutual information and domain-transformed wideband deconvolution.First,an explicit convolution model for wideband signals in the delay-scale domain is derived.Then,a scale factor pre-estimation method based on matrix mutual information is introduced to reduce the dimensionality of the deconvolution process.Finally,the Richardson-Lucy iterative algorithm is employed to achieve high-resolution reconstruction of the target reflection function.Simulation results demonstrate that the proposed method maintains high robustness in scale estimation under low signal-to-noise ratio conditions,improves delay resolution for echoes from multiple scatterers,and effectively suppresses sidelobes.Experimental results further validate the effectiveness and feasibility of the method in achieving high-resolution target parameter estimation.This provides an effective approach for weak target detection and feature extraction in complex underwater acoustic environments.
In order to solve the problem of low audio fingerprint retrieval recognition rate under background sound and noise conditions,a novel algorithm based on mute masking and frequency segmentation is proposed.In the fingerprint extraction stage,voice activity detection technology is used to remove the non-valid speech frames,and then the valid speech frames are recombined and features are extracted according to the difference of the adjacent sub-band energy,which can effectively solve the problem that silence frame fingerprint characteristics are not robust.In the matching stage,according to the distribution characteristics of different audio signals in the frequency domain,the audio fingerprints are segmented and weighted in different frequencies to calculate the similarity between the template and the test audio more accurately.Experiments show that the proposed algorithm doubles the retrieval speed compared with the classic Philips algorithm.In the meantime,it yields a large definite improvement over Philips by 17.94％on mean average precision and 4.66％on recall respectively for the data set disturbed by background sounds.Compared with the latest Philips algorithm,the mean average precision and recall have been increased by 13.68％and 2.45％respectively.
Under low signal-to-noise ratio(SNR)and burst noise conditions,the speech en-hancement effect of existing deep learning network models is not satisfactory.In contrast,humans can exploit the long-term correlation of speech to form an integrated perception of dif-ferent speech signals.Thus,describing the long-term dependencies of speech can help improve the enhancement performance under low SNR and burst noise conditions.Inspired by this feature,a time domain end-to-end monaural speech enhancement model TU-net that fuses the multi-head self-attention mechanism and U-net deep network is proposed.The TU-net model adopts the codec layer structure of U-net to achieve multi-scale feature fusion.It introduces the dual-path Transformer module using the multi-head self-attention mechanism to calculate the speech mask and better model long-term correlation.The TU-net model is trained with a weighted sum loss function in the time,time-frequency,and perceptual domains.Simulation experiments are carried out and the results show that with maintaining relatively fewer network model parameters,TU-net outperforms other similar monaural enhancement network models in several evaluation metrics such as perceptual evaluation of speech quality(PESQ),short-time objective intelligibility(STOI)and SNR gain under low SNR and burst noise conditions.
A*Orthogonal Matching Pursuit(A*OMP)was modified and applied to underwa-ter acoustic channel estimation,which outperformed than OMP.A*OMP is a pseudo-greedy algorithm,and can improve the problem of OMP algorithm falling into local optimization.Even though A*OMP has better performance than OMP,the channel estimation error can-not decrease as signal-to-noise ratio(SNR)increases.To solve the problem,the initialization procedure and termination criterion of A*OMP was modified for channel estimation,to select right initial searching path and avoid too many iterations which cause unknown errors.Based on orthogonal frequency division multiplexing(OFDM)system,numerical simulation results showed that the pilot interval did not obviously affect the performance of the modified A*OMP.And there was no channel estimation error floor for the modified A* OMP.The error of the pro-posed algorithm was reduced by two and one orders of magnitude respectively from those of OMP and A*OMP.The proposed method outperformed the OMP and traditional A*OMP algorithm in both estimation accuracy and communication performance,with bit error rate(BER)reductions of 42.0％and 4.7％,respectively.
Active sonar can separate clutter,reverberation,and moving targets in the Doppler frequency shift domain using Doppler sensitive signals,but time and Doppler leakages of strong interference can inundate weak targets at low signal-to-interference ratios.Therefore,a small moving target interference suppression detection method based on an adaptive least mean square(LMS)algorithm and wide-band ambiguity function(WAF)is proposed.First,an adap-tive notch filter based on LMS is used to suppress interference in the spatial Doppler frequency shift domain and then WAF used to detect targets in the fast-time Doppler frequency shift do-main.Numerical simulation and pool experiments are performed and the results demonstrated that the proposed method effectively suppress strong interference and accurately estimate the target delay and Doppler shift.About 13 dB of interference suppression is observed in exper-iments.In addition,the system is able to tolerate interference fluctuations with normalized amplitude fluctuation variance less than 0.5 and improve the active detection capacity for small moving targets.
A piezoelectric planar transducer with high sensitivity is designed and fabricated.The transducer sensing element is an improvement of the 1-3-2 type piezoelectric composite structure,the piezoelectric ceramic column array with a substrate is not injected with poly-mer,and the upper surface is directly covered with a metal plate,the improved structure of"piezoelectric column array with substrate+metal cover plate"sensitive element(called"Air filled"sensitive element)is formed.The resonant frequency of the"Air filled"transducer is calculated theoretically and simulated by the finite element method,which is in good agree-ment with the measured result.In order to compare the performances,the transducer with the structure of"1-3-2 type piezoelectric composite+metal cover plate"(called"Polymer filled"transducer)is made.The effective electromechanical coupling coefficient,transmitting voltage response and receiving sensitivity of"Air filled"and"Polymer filled"transducers are simulated and measured respectively.The"Air filled"transducer has a higher receiving sensitivity,which can be increased by 21 dB compared with the"Polymer filled"transducer.The transducer can effectively improve the sensitivity,which can provide a reference for the development of high-sensitivity transducer.
A potential risk in ultrasonic guided wave testing is that weak echo signals from small defects may be submerged in noisy signals,which will cause missed detection.To over-come this shortcoming,a weak guided wave signal detection method based on period jump of the Duffing system is proposed in this paper.The critical state of the system period jump can be obtained by analyzing the bifurcation characteristics of the Duffing system with the variation of the driving force amplitude.A weak ultrasonic guided wave signal with the same frequency as the driving force is added to the driving force.This is equivalent to changing the driving force amplitude,which causes the period state to jump.Consequently,the weak guided wave signals can be identified based on the period jumps.The increase or decrease in the driving force amplitude due to the interference of the guided wave signal depends on the phase difference between the intercepted signal and the periodic driving force.The conditions for increasing and decreasing the driving force amplitude are out of phase with each other.They have an approximate phase difference of π within the same period.Two detection models for small-scale periodic states (SPS) and large-scale periodic states (LPS) are constructed,and the effectiveness of the models in identifying the guided wave signal is verified numerically and experimentally.The anti-noise interference capabilities of the two models are also compared.The results show that the SPS detection model provides unique results and a strong anti-noise ability,and effectively improves the sensitivity of small defect detection using ultrasonic guided waves.
Echolocation clicks of Indo-Pacific humpback dolphins(Sousa chinensis)and East Asian finless porpoises(Neophocaena asiaeorientalis sunmeri)were recorded and analyzed to deepen our understandings on vocalizations variability of these two endangered species.Passive acoustic monitoring was used to record the clicks during boat-based field surveys in the Xia-men waters.Subsequent analyses show clicks emitted by Indo-Pacific humpback dolphins are characterized by short duration,high frequency,and broadband bandwidth.Comparatively,clicks of East Asian finless porpoise have a relatively longer duration,higher peak frequency,and narrower bandwidth.The duration,centroid frequency and-10 dB bandwidth of clicks are significantly different between these two species.
An improved U-Net(Attention Dilated Convolution U-Net,ADC-U-Net)network model for end-to-end speech enhancement is designed based on the U-Net network.Compared with the baseline U-Net network,the dilated convolution is added to reduce the loss of infor-mation caused by sampling.Besides,the attention mechanism structure is introduced,which combines more contextual information of noisy speech to extract deeper and richer feature in-formation.The proposed model avoids the extraction of features with distinct step,so it does not need three steps of traditional methods,including feature extraction,feature denoising and waveform reconstruction.The proposed network model obtains complex structural features to represent speech through multi-level and multi-scale learning.The quality and intelligibility of enhanced speech are evaluated by several subjective and objective indexes.Experimental data show that the proposed algorithm performs well in noise suppression and adaptability,and has advantage over baseline U-Net network and other models in speech quality and intelligibility.
Aiming at the problem that the supervised speech enhancement ignores the impact of the similarity of amplitude spectrum between clean speech,noise,and noisy speech on the enhancement effect,a method combining accurate ratio mask(ARM)and deep neural network(DNN)is proposed for monaural speech enhancement.Firstly,an accurate ratio mask based on ideal ratio mask in the time-frequency domain is designed,which utilizes the normalized cross-correlation coefficients of amplitude spectrum between clean speech and noisy speech,and between noise and noisy speech.Then,the target mask is estimated by the output of the baseline DNN which takes the amplitude spectrum of clean speech and noise as training target,and further uses the target mask to optimize the baseline DNN and gets the enhanced speech from noisy speech.Moreover,considering the discriminative information between clean speech and noise,a discriminative training function is used to replace the mean square error(MSE)as the objective function of the DNN,thus making the output of network more accurate.The experimental results show that the discriminative training function improves the enhancement effect of baseline DNN and the overall joint optimization network.Compared with other com-mon DNN methods,the proposed method has achieved higher average PESQ,STOI and better noise suppression effect under matched and unmatched noise,and the enhanced speech also retains more speech components.
To investigate the noise issue caused by open-cavity self-sustained oscillations,flow over a rectangular cavity with a length-to-depth ratio of 2 at Mach numbers of 0.10,0.15,0.20,and 0.25 was experimentally studied in a 0.55 m×0.40 m aeroacoustics wind tunnel.Further,the velocity field and self-sustained oscillation were studied.High-frequency particle image velocimetry was employed to visualise the flow-field structure and analyse the path of the acoustic and pressure waves.The acoustic self-sustaining oscillation modes and duct modes were investigated by testing the far field and the fluctuating wall pressure.Moreover,the characteristics of the correlation between the wall pressure and the acoustic performance were discussed.The results demonstrated that apart from the main vortex structure,shear vortices at the leading edge and crash vortices at the trailing edge were observed.Furthermore,around the frequencies of 875 Hz,1288 Hz,1875 Hz,and 2050 Hz,an evident tonal noise,induced by the cavity oscillation,could be detected.In addition,the wall pressure strongly correlated with the far-field noise,and the tone appeared at the same frequencies where the peaks occurred in the wall pressure spectra.
A theoretical model of a propeller cavitation noise modulation spectrum of ships with skewed propellers is proposed.According to the relationships between the wake velocity and the cavitation noise volume and between the propeller rotational speed and the number of bubbles,a mathematical expression of the cavitation noise modulation spectrum is derived.Theoretical analysis and simulation show that the difference between the blades is the main reason for the existence of a shaft frequency line spectrum in the modulation spectrum and that the main affecting factors of the modulation spectrum intensity of blade frequency in the modulation spectrum are the propeller's skew angle and revolution.The results presented in this study provide a theoretical basis for the analysis and extraction of modulation spectrum characteristics of the radiated noise and also have a reference value for the low-noise structure design of ship propellers.
In deep ocean environments,the bottom bounce mode associated with the low-frequency active sonar is often used to detect the underwater target located in the first shadow zone.However,range estimation error occurs due to the varying sound speed,which produces a bending acoustic ray path from the sonar to the target.To reduce the active ranging error,a method that uses the effective sound speed can be used.However,this method faces a heavy computation burden due to the calculation of the time delay-effective sound speed pairs at each grid position in the bottom bounce area.To reduce this computation burden,an improved effective sound speed estimation method based on the interference structure of the sound field is proposed.The sound field fluctuation in the bottom bounce area is due to the energy fluctuation of the sound rays having different grazing angles.A quantitative relationship between the detectable areas and the sound rays with high energies is established based on sound ray interference theory.The time delay-effective sound speed pairs in the boundary location of the detectable areas are calculated according to this relationship.Hence,the time delay-effective sound speed pairs at all the grid positions can be obtained by linear fitting.Simulation results show that the improved method can obtain a similar range estimation error as the conventional effective sound speed estimation method but has a much lower computation burden,which is useful for real-time applications.
The signal-to-direct-blast ratio SDRF in acoustic forward-scattering detection can measure the relative magnitudes of the forward scattered wave and the direct-blast,but it does not consider the interference and superimposition effects between the two waves.These two waves inevitably interfere with one another and are difficult to distinguish,so it is difficult to directly apply SDRF in target detection and analysis.Based on SDRF and considering the interference between the forward scattered wave and the direct-blast,a new parameter called the acoustic interfered field distortion ΔFTL is developed,and the corresponding calculation formula is deduced.Compared with SDRF,ΔFTL can be obtained directly from data and has the advantage of not relying on prior information to a certain extent.The following is found according to this formula in combination with data from the Qiandao Lake scaled-target detection experiment.(1)Estimating the geometric expansion loss coefficient reveals that the acoustic wave propagates spherically,which is consistent with the simulation results of the ray model based on the measured hydrological parameters.(2)The relation between ΔFTL and the target crossing position is quantitatively confirmed,confirming the effectiveness of the ΔFTL formula.(3)A performance evaluation scheme independent of prior information is established for direct-blast suppression and then applied to an adaptive direct-blast suppression method.The influences of acoustic leakage and disturbances in the attitude angle of the target on the results can be ignored.These results show that ΔFTL can effectively replace SDRF,providing theoretical references for conducting performance evaluations on acoustic forward-scattering detection and direct-blast suppression.
