Background Noise Research Articles

Distributed estimation cascade second-order tri-stable stochastic resonance method for random noise suppression in continuous-wave mud pulse telemetry

The extraction of high-quality useful signals in downhole environments has perennially posed a technical challenge for continuous wave mud pulse transmission systems. Traditional denoising methods, such as adaptive filtering and wavelet transform, inevitably compromise the integrity of the original signal's useful components while removing noise. Different from the traditional denoising methods, stochastic resonance (SR) is capable of transferring the energy of noise to the signal, thereby accomplishing the objective of noise suppression. In this paper, based on the second-order tri-stable stochastic resonance detection method combined with the estimation of distribution algorithm (EDA), a distributed estimation cascade second-order tri-stable stochastic resonance (EDA- CSTSR) detection method is proposed for the first time. By constructing 5 groups of simulation signals with different signal-to-noise ratios (SNRs), the denoising performance of distributed estimation CSTSR, Langevin SR, and Duffing SR is analyzed in detail. The results show that compared with Langevin SR and Duffing SR, the EDA-CSTSR can increase the SNR of the input simulation signal by at least 17 dB and significantly increase the amplitude of the useful pulse signal. Moreover, the EDA-CSTSR is applied to process actual data obtained from a drilling site, enabling the recognition of weak signals amidst strong background noise, which highlights the method's excellent practical application value and underscores its potential for further enhancement.

Objective and subjective evaluation of speech enhancement methods in the UDASE task of the 7th CHiME challenge

Supervised models for speech enhancement are trained using artificially generated mixtures of clean speech and noise signals. However, the synthetic training conditions may not accurately reflect real-world conditions encountered during testing. This discrepancy can result in poor performance when the test domain significantly differs from the synthetic training domain. To tackle this issue, the UDASE task of the 7th CHiME challenge aimed to leverage real-world noisy speech recordings from the test domain for unsupervised domain adaptation of speech enhancement models. Specifically, this test domain corresponds to the CHiME-5 dataset, characterized by real multi-speaker and conversational speech recordings made in noisy and reverberant domestic environments, for which ground-truth clean speech signals are not available. In this paper, we present the objective and subjective evaluations of the systems that were submitted to the CHiME-7 UDASE task, and we provide an analysis of the results. This analysis reveals a limited correlation between subjective ratings and several supervised nonintrusive performance metrics recently proposed for speech enhancement. Conversely, the results suggest that more traditional intrusive objective metrics can be used for in-domain performance evaluation using the reverberant LibriCHiME-5 dataset developed for the challenge. The subjective evaluation indicates that all systems successfully reduced the background noise, but always at the expense of increased distortion. Out of the four speech enhancement methods evaluated subjectively, only one demonstrated an improvement in overall quality compared to the unprocessed noisy speech, highlighting the difficulty of the task. The tools and audio material created for the CHiME-7 UDASE task are shared with the community.

Advancements in hand-drawn chemical structure recognition through an enhanced DECIMER architecture

Accurate recognition of hand-drawn chemical structures is crucial for digitising hand-written chemical information in traditional laboratory notebooks or facilitating stylus-based structure entry on tablets or smartphones. However, the inherent variability in hand-drawn structures poses challenges for existing Optical Chemical Structure Recognition (OCSR) software. To address this, we present an enhanced Deep lEarning for Chemical ImagE Recognition (DECIMER) architecture that leverages a combination of Convolutional Neural Networks (CNNs) and Transformers to improve the recognition of hand-drawn chemical structures. The model incorporates an EfficientNetV2 CNN encoder that extracts features from hand-drawn images, followed by a Transformer decoder that converts the extracted features into Simplified Molecular Input Line Entry System (SMILES) strings. Our models were trained using synthetic hand-drawn images generated by RanDepict, a tool for depicting chemical structures with different style elements. A benchmark was performed using a real-world dataset of hand-drawn chemical structures to evaluate the model's performance. The results indicate that our improved DECIMER architecture exhibits a significantly enhanced recognition accuracy compared to other approaches.Scientific contributionThe new DECIMER model presented here refines our previous research efforts and is currently the only open-source model tailored specifically for the recognition of hand-drawn chemical structures. The enhanced model performs better in handling variations in handwriting styles, line thicknesses, and background noise, making it suitable for real-world applications. The DECIMER hand-drawn structure recognition model and its source code have been made available as an open-source package under a permissive license.Graphical

Performance and detection range of acoustic receivers in mangrove habitats.

Acoustic telemetry has been used to monitor the movement of aquatic animals in a broad range of aquatic environments. Despite their importance, mangrove habitats are understudied for the spatial ecology of elasmobranchs, with acoustic telemetry rarely used inside mangrove habitats. One reason for this may be a general assumption that acoustic signals would not be able to be detected by receivers in such shallow, structurally complex, environments. This study tested whether acoustic receivers can be used inside mangrove habitats to track the movement of sharks and rays. Thirty-eight receivers were deployed in a mangrove system in Pioneer Bay, Orpheus Island, Great Barrier Reef, including inside mangroves, mangrove edges, and adjacent reef flat areas. The detection range and receiver performance metrics, such as code detection efficiency, rejection coefficient, and noise quotient, were examined and tested among habitats. The results highlighted that the signal from transmitters was successfully detected inside mangrove habitats as well as on the adjacent reef flat. The range to detect at least 50% of transmissions was up to 20 m inside mangroves and up to 120 m outside mangroves. The performance metrics of acoustic receivers inside the mangrove habitat were characterized by low background noise, low rejection rates, and reasonably high code detection efficiency. Furthermore, this study tested the application of this method on juvenile blacktip reef shark Carcharhinus melanopterus and mangrove whipray Urogymnus granulatus, and demonstrated that it can be used to successfully track animals inside mangrove habitat. This novel method could reveal further information on how sharks and rays use mangrove habitats.

DACBN: Dual attention convolutional broad network for fine-grained visual recognition

Fine-grained visual classification (FGVC) is a challenging task due to its small inter-class differences and large intra-class differences. Most existing methods rely on manual labeling of key identification areas, which requires high labor costs. In addition, existing methods also tend to ignore the differences effect of different feature channels in the feature map, which has a certain impact on the model classification accuracy. To solve the above problems, this paper proposes a dual attention convolutional broad network. Firstly, a new dual attention mechanism is designed to suppress the background noise of fine-grained images and give greater weight to the discriminative feature regions and channels. Secondly, the ensemble broad learning system framework is used to further enhance the dual attention features, so that the discriminative features can further improve the recognition ability of the model. Finally, by multiple comparative experiments, it is reported that the method proposed in this article has achieved excellent recognition results on three commonly used datasets.

Hou Hsiao-Hsien’s acousmatic sound: Listening to urban noises and decadent music in Daughter of the Nile

This paper reconsiders the mediation of urban noises and decadent music in Hou Hsiao-hsien’s long-overlooked film Daughter of the Nile (1987) under the contexts of noise control and popular entertainment in the late 1980s in Taiwan. It argues that the film embeds distinctive sounds in urban environments to create acousmatic soundscapes that are typically considered ambient noises in mainstream discourse. It further observes that the film provides a critical entry point to the historical debates over the definition and gov­ernance of noises at the time, complicating an overdetermined ­concept throughout the history of sonic perception. Offering a frame­work to understand the film’s mediation of Taipei’s urban soundscapes, the paper suggests that the film implies an ethic of listening that calls for a more careful attention to the multifarious sounds that reflect the power dynamics between the authority and the residents in the city.

Depth-resolved imaging through dynamic scattering media via speckle cross-correlation under near-infrared illumination

Speckle cross-correlation imaging (SCCI) method has the depth-resolved capability, benefiting from the introduction of a reference point. However, the quality of the reconstructed image is degraded due to the background noise, which becomes more prominent when imaging through dynamic scattering media. Here, we propose a composite-differential filter-assisted (CDF-SCCI) method, allowing for effectively reducing the background noise of the reconstructed image. The signal-to-background ratios of the reconstructed images by employing the CDF-SCCI method can be enhanced by a maximum of 4.15 dB (corresponding to 2.6 times) compared to the SCCI method. Furthermore, we apply the near-infrared (NIR) illumination to the imaging system of dynamic scattering media and prove that the NIR illumination not only enhances the penetration depth of imaging, but also improves the quality of reconstructed images compared to the visible illumination. The depth-resolved imaging through various dynamic biological scattering media, including the milk and anticoagulated pig blood, further demonstrates the potential application of the proposed CDF-SCCI method in biomedical imaging.

PMANet: Progressive multi-stage attention networks for skin disease classification

Automated skin disease classification is crucial for the timely diagnosis of skin lesions. However, accurate skin disease classification presents a challenge, given the significant intra-class variation and inter-class similarity among different kinds of skin diseases. Previous studies have attempted to address this issue by identifying the most discriminative part of a lesion, but they tend to overlook the interactions between multi-scale features. In this paper, we propose a Progressive Multi-stage Attention Network (PMANet) to enhance the learning of multi-scale discriminative features, so that the model can gradually localize from stable fine-grained to coarse-grained regions in order to improve the accuracy of disease classification. Specifically, we utilize a progressive multi-stage network to supervise feature and classification, thereby fostering multi-scale information and improving the model's ability to learn intra-class consistent information. Additionally, we propose an enhanced region proposal block that highlights key discriminative features and suppresses background noise of lesions, reinforcing the learning of inter-class discriminative features. Furthermore, we propose a multi-branch feature fusion block that effectively fuses multi-scale lesion features from different stages. Comprehensive experiments conducted on two datasets substantiate the effectiveness and superiority of the proposed method in accurately classifying skin disease.

Improving signal-to-noise ratios of ambient noise cross-correlation functions using local attributes

SUMMARY For seismographic stations with short acquisition duration, the signal-to-noise ratios (SNRs) of ambient noise cross-correlation functions (CCFs) are typically low, preventing us from accurately measuring surface wave dispersion curves or waveform characteristics. In addition, with noisy CCFs, it is difficult to extract relatively weak signals such as body waves. In this study, we propose to use local attributes to improve the SNRs of ambient noise CCFs, which allows us to enhance the quality of CCFs for stations with limited acquisition duration. Two local attributes: local cross-correlation and local similarity, are used in this study. The local cross-correlation allows us to extend the dimensionality of daily CCFs with computational costs similar to global cross-correlation. Taking advantage of this extended dimensionality, the local similarity is then used to measure non-stationary similarity between the extended daily CCFs with a reference trace, which enables us to design better stacking weights to enhance coherent features and attenuate incoherent background noises. Ambient noise recorded by several broad-band stations from the USArray in North Texas and Oklahoma, the Superior Province Rifting EarthScope Experiment in Minnesota and Wisconsin and a high-frequency nodal array deployed in the northern Los Angeles basin are used to demonstrate the performance of the proposed approach for improving the SNR of CCFs.

Uncertainty Quantification in Radial Anisotropy Models Based on Transdimensional Bayesian Inversion of Receiver Functions and Surface-Wave Dispersion: Case Study Sri Lanka

ABSTRACT In geophysical inference problems, quantification of data uncertainties is required to balance the data-fitting ability of the model and its complexity. The transdimensional hierarchical Bayesian approach is a powerful tool to evaluate the level of uncertainty and determine the complexity of the model by treating data errors and model dimensions as unknown. In this article, we take account of the uncertainty through the whole procedure, thus developing a two-step fully Bayesian approach with coupled uncertainty propagation to estimate the crustal isotropic and radial anisotropy (RA) model based on Rayleigh and Love dispersion as well as receiver functions (RFs). First, 2D surface-wave tomography is applied to determine period-wise ambient noise phase velocity maps and their uncertainty for Rayleigh and Love waves. Probabilistic profiles of the isotropic average VS and RA as a function of depth are then derived at station sites by inverting the local surface-wave dispersion and model errors and RFs jointly. The workflow is applied to a temporary seismic broadband array covering all of Sri Lanka. The probabilistic results enable us to effectively quantify the uncertainty of the final RA model and provide robust inferences. The shear-wave velocity results show that the range of Moho depths is between 30 and 40 km, with the thickest crust (38–40 km) beneath the central Highland Complex. Positive RA (VSH>VSV) observed in the upper crust is attributed to subhorizontal alignment of metamorphic foliation and stretched layers resulting from deformation. Negative RA (VSV>VSH) in the midcrust of central Sri Lanka may indicate the existence of melt inclusions and could result from the uplift and folding process. The positive RA in the lower crust could be caused by crustal channel flow in a collision orogeny.

Listening and learning challenges for deaf or hard of hearing students during different classroom scenarios: an itinerant teachers of the deaf perspective

ABSTRACT The aim of this exploratory study was to gather survey data from itinerant teachers of the deaf (IToDs) to better understand the impact of different classroom scenarios on 6- to 9-year-old Deaf or hard of hearing students’ listening and learning. Across two surveys, 109 IToDs (Study 1) and 18 IToDs (Study 2) were either asked to list the positive or negative implications for students’ listening and learning during lecture-style lessons, guided-learning groups, and self-directed/independent work (Study 1), or were asked to rank the importance of each of the implications (Study 2). The different types of classroom scenarios differentially impacted students’ listening and learning outcomes. In lecture-style, the IToDs reported that assistive devices worked well, but there was less support for the student in this scenario. For guided-learning, IToDs identified background noise as problematic, but as the students were engaged in classroom discussions, the IToDs felt that it was also a beneficial scenario for developing social skills. When students were engaged in self-directed/independent work, IToDs noted that while the quieter environment could be beneficial for the student’s concentration, it can also be quite isolating for the student. Overall, this study provided helpful insights from IToDs into the factors affecting listening and learning during different classroom scenarios for students who are Deaf or hard of hearing. The challenges and benefits these students encounter in the different learning settings revealed valuable insights that can be utilized by educators and policymakers aiming to optimize educational outcomes for this demographic.

Integrated Amphibious Distributed Acoustic Sensing for Seismic Monitoring in the Xinfengjiang Reservoir

Abstract The rapidly advancing technology of distributed acoustic sensing (DAS) has profoundly impacted the field of underwater geophysics. Our study investigates the effectiveness of DAS in underwater geological stability monitoring, with a particular focus on microseismic monitoring in the Xinfengjiang reservoir. The 6.2 km long acquisition setup, covering both land and reservoir bottom, was verified using temporary shore-based short-period seismometers to ensure reliable data acquisition in various environments. Higher background noise was observed on the land section compared with the lakebed section during the day, whereas both sections exhibited similar noise levels at night. We confirmed that the DAS system was capable of detecting distant microseismic events, some of which were previously unreported. These detections exhibited temporal and phase consistency with neighboring seismometers. Comparison of signal-to-noise ratios indicates that the lakebed section demonstrates higher sensitivity. This system delivers cost-effective performance through natural settling, negating the requirement for costly embedding methods. Moreover, the DAS system identified “comet-like” small-scale signals on the lakebed that had eluded shore-based seismometers. This exemplifies the exceptional high-density and high-resolution capabilities of DAS technology in both aquatic and terrestrial environments. This study underscores the pivotal role of the DAS technology in conducting underwater microseismic monitoring, real-time seismic monitoring, seismic mechanism research, and earthquake hazard assessment.

Text-independent voiceprint recognition via compact embedding of dilated deep convolutional neural networks

In order to process speech, most state-of-the-art experimental methods employ convolutional neural networks (CNNs), which operate on a continuous, 1-dimensional (1-D) time stream. In an audio signal, the mel-spectrogram facilitates the representation of attributes of the utterances' in the frequency domain (which corresponds to the speech spectrum). Moreover, for a time-series speaker signal, CNNs are superior to machine or transfer learning models in capturing characteristics from long-form talks. This paper introduces a jump-connected 1-D CNN that employs a combined loss function for speaker recognition. The suggested model uses a 1-D convolutional layer combined with jump connections to extract speaker-specific characteristics; this reduces time-based and frequency-based variability for faster computing. A combined softmax loss, stable L2-norm, and smooth L1-norm loss function guide the proposed compact convolutional neural networks (CCNN) to identify the correct spokesman with improved efficacy. We evaluated the proposed framework using various standard and real-time audio datasets. The experimental findings demonstrate that the proposed CCNN outperforms existing approaches by reducing the equal error rate by 9.02 %. Also, our recommended voiceprint identification model achieves an impressive average speaker recognition rate of 98.76 %. Simultaneously, the reliability of the 1-D CCNN is tested under various conditions. Other fields of study, like language modelling, could employ this approach after some fine-tuning.Relevance of the work: Speaker recognition is an area of interest in which machine learning (ML) and deep learning (DL) schemes, when combined, have the potential to make history in the areas of forensic sciences, automation, and authentication. Using a modest CNN can enhance the identification and verification process by ignoring many issues such as false positives, background noise, and so on. Expanding this process would facilitate raga identification and disease treatment therapies.

Fusing crops representation into snippet via mutual learning for weakly supervised surveillance anomaly detection

AbstractIn recent years, the challenge of detecting anomalies in real‐world surveillance videos using weakly supervised data has emerged. Traditional methods, utilising multi‐instance learning (MIL) with video snippets, struggle with background noise and tend to overlook subtle anomalies. To tackle this, the authors propose a novel approach that crops snippets to create multiple instances with less noise, separately evaluates them and then fuses these evaluations for more precise anomaly detection. This method, however, leads to higher computational demands, especially during inference. Addressing this, our solution employs mutual learning to guide snippet feature training using these low‐noise crops. The authors integrate multiple instance learning (MIL) for the primary task with snippets as inputs and multiple‐multiple instance learning (MMIL) for an auxiliary task with crops during training. The authors’ approach ensures consistent multi‐instance results in both tasks and incorporates a temporal activation mutual learning module (TAML) for aligning temporal anomaly activations between snippets and crops, improving the overall quality of snippet representations. Additionally, a snippet feature discrimination enhancement module (SFDE) refines the snippet features further. Tested across various datasets, the authors’ method shows remarkable performance, notably achieving a frame‐level AUC of 85.78% on the UCF‐Crime dataset, while reducing computational costs.

Enhanced Epileptic Seizure Detection through Wavelet-Based Analysis of EEG Signal Processing

Epilepsy affects millions worldwide, making timely seizure detection crucial for effective treatment and enhanced well-being. Electroencephalogram (EEG) analysis offers a non-intrusive solution, but its visual interpretation is prone to errors and requires a lot of time. Many existing works focus solely on achieving competitive levels of accuracy without considering processing speed or the computational complexity of their models. This study aimed to develop an automated technique for identifying epileptic seizures in EEG data through analysis methods. The efforts have been primarily focused on achieving high accuracy results by operating exclusively within a narrow frequency band of the signal, while also aiming to minimize computational complexity. In this article, a new automated approach is presented for seizure detection by combining signal processing and machine learning techniques. The proposed method comprises four stages: (1) Preprocessing: Savitzky–Golay filter to remove the background noise. (2) Decomposition: discrete wavelet transform (DWT) to extract spontaneous alpha and beta frequency bands. (3) Feature extraction: six features (mean, standard deviation, skewness, kurtosis, energy, and entropy) are computed for each frequency band. (4) Classification: a support vector machine (SVM) method classifies signals as normal or containing a seizure. The method was assessed using two publicly available EEG datasets. For the alpha band, the highest achieved accuracy was 92.82%, and for the beta band it was 90.55%, which demonstrates adequate capability in both bands for accurate seizure detection. Furthermore, the obtained low computational cost suggests a potentially valuable application in real-time assessment scenarios. The obtained results indicate its capacity as a valuable instrument for diagnosing epilepsy and monitoring patients. Further research is necessary for clinical validation and potential real-time deployment.

Unique molecular identifier-based amplicon sequencing of microhaplotypes for background noise mitigation

Microhaplotypes (MHs), comprising two or more single-nucleotide polymorphisms in a short fragment, are promising forensic markers owing to their remarkable polymorphic nature. Several studies have demonstrated the utility of MHs through massively parallel sequencing (MPS). Nevertheless, the background noise level associated with MHs in MPS, which imposes a practical detection limit for the system, remains uninvestigated. Currently, unique molecular identifier (UMI) systems are known to effectively mitigate background noise by tracking original DNA molecules and facilitating PCR and MPS error corrections. Hence, this study aimed to design a UMI-based amplicon sequencing system, designated MH-UMIseq, which can amplify 46 MHs simultaneously and generate MPS libraries in four steps: barcoding PCR, nuclease reaction, boosting PCR, and indexing PCR. The performance of the MH-UMIseq system was evaluated using the Illumina NextSeq 550 and MiniSeq systems with 31 sets for 5 ng, 1 ng, and 200 pg of input DNA. The fgbio toolkit was used in conjunction with STRait Razor 3.0 and Visual Microhap to analyze the UMI data on MHs. The corresponding average not suppressed noise proportion of MH-UMIseq were 0.1 %, 0.3 %, and 0.7 % for 5 ng, 1 ng, and 200 pg of DNA, respectively, which substantially suppressed the background noise for more than 1 ng of DNA. Interestingly, the proportion of not suppressed noise in MH-UMIseq notably decreased as the amount of input DNA increased. The number of UMI families was proportional to the copy number of the template DNA and closely correlated with the system resolution. Therefore, the resolution of MH-UMIseq system is expected to be higher than that of conventional MPS for the deconvolution of mixtures containing more than 1 ng of DNA.

3-D Shear wave velocity and azimuthal anisotropy structure in the shallow crust of Binchuan Basin in Yunnan, SW China, from ambient noise tomography

The Binchuan Basin in northwest Yunnan, SW China, is a rift basin developed at the intersection of the Red River Fault and Chenghai Fault, where active faults developed and historical earthquakes occurred. Understanding the fine velocity structure of the shallow crust in this region is crucial for comprehending earthquake occurrences and mechanisms and for improving earthquake location accuracy. Using the continuous waveform data recorded by 381 dense array stations in 2017, we obtained 7915 Rayleigh-wave phase velocity dispersion curves in the period band of 0.2-6 s from ambient noise cross-correlation functions after rigorous data processing and quality control. We determined 3-D isotropic and azimuthally anisotropic shear wave velocity models at depths above 6 km in the shallow crust based on the direct surface wave azimuthal anisotropic tomography method. The isotropic model reveals a strong correspondence between the S-wave velocity structure at depths of 0-1 km and the regional topography and lithology. The Binchuan depocenter, Zhoucheng depocenter, Xiangyun Basin, and Xihai Rift Basin are primarily composed of Quaternary deposits, which show low-velocity anomalies, while the regions with the Paleozoic shale, limestone, and basalt exhibit high-velocity anomalies. The nearly N-S orientation of fast directions from azimuthal anisotropy models are mainly controlled by the active Binchuan Fault with N-S strike as well as the NNW oriented primary compressive stress.

Site Characterization and Shallow Shear Velocity through HVSR Measurements around the Hyderabad Metropolitan Region, India

ABSTRACT We investigate the site characterization and shallow shear velocity profiles from the analysis of the Horizontal to Vertical Spectral ratio (HVSR) around the Hyderabad metropolitan region (HMR), which falls under the southern Indian shield. This work uses both the ambient noise and microearthquake data to compute the HVSR, and additionally the Random Decrement technique to compute the HVSR of extracted Rayleigh waves. This study indicates comparable HVSR curves at each station with the three different datasets, from which we obtain the average dominant frequency (f0) and amplification value (A0). They are further used to calculate the seismic vulnerability index value (Kg). We observe that the value of f0 around the HMR is not fixed, but is varying in the range of 3.4 to 18 Hz, whereas the value of A0 is in the range of 1.7 to 12 approximately and Kg in the range of 0.16 to 1.68 approximately, with an exception of ∼ 33 at VKB (Vikarabad) station, which may be due to a local unconsolidated sub-surface structure. Based on the Diffused Field Assumption (DFA), we invert the average HVSR curves and average dispersion curves of Rayleigh waves, and provide the shallow shear velocity profiles up to 300 m, along with an approximate estimate of VS30 (in upper 30 m depth). The estimated VS30 values vary between 911 to 3143 m/s, falling under the classifications A and B (mostly Hard Rock and Rock type) of National Earthquake Hazards Reduction Program (NEHRP) (BSSC, 2001). However, our study shows some stations with shear velocity inversions at shallow depths within 300 m, indicating layers of low velocity, needing further study. In the absence of detailed near-surface findings, these findings are valuable inputs for geotechnical engineering studies and urban-city planning around the HMR, and emphasizes the effectiveness the HVSR method to determine sub-surface topography and/or unknown soil structures as an economical investigation viability.

Crustal structure beneath the Western Dharwar Craton segment of Western Ghats: Insights from ambient noise correlation technique

The Western Ghats (WG) comprise three major tectonic blocks: the Deccan Volcanic Province (DVP), Western Dharwar Craton (WDC) and Southern Granulite Terrain (SGT) from north to south. Among these, the WDC region encompasses the oldest basement rocks, including greenstone belts and calc-alkaline plutons dating back to ∼3.0 to 3.4 Ga. While the crustal structure of the DVP and SGT segments of the WG has been extensively studied, knowledge regarding the crustal structure beneath the WDC segment of the WG is limited due to the scarcity of seismic stations. To address this gap, we utilized continuous waveform data from a newly established network of seismic stations by the National Centre for Earth Science Studies (NCESS), in addition to other available stations and deciphered the shallow crustal shear-wave velocity structure along a profile beneath the WDC segment of the WG using ambient noise cross-correlation and surface wave dispersion. The results reveal a continuous upper layer characterized by a high shear wave velocity of ∼3.6 km/s with notable variation in thickness along the profile. Furthermore, a thick middle layer with velocities reaching up to ∼3.9 km/s and a lower layer with velocities of ≥4.2 km/s are observed. Intriguingly, variations in the thickness of the upper layer, ranging from 5 km to 20 km, are noted beneath the study region. The thickness of this layer exhibits changes in relation to topographical variations, with this dependency diminishing for deeper layers. Notably, high shear wave velocity perturbations (∼10 %) are observed in the upper layer beneath low-elevation regions, possibly indicating the presence of Tonalite–Trondhjemite–Granodiorite (TTG)-type gneiss − a significant rock type in the WDC region, alongside greenstone belts and late calc-alkaline to potassic plutons. These findings suggest that the uneven topography of the WG may result from long-term erosion, exposing deeper and higher-density rocks and leading to higher shear wave velocities at shallow depths. This study serves as a foundation for further exploration into the influence of the upper crustal layer and its correlation with the evolution of the WG.

Machine-learning based detection of marine mammal vocalizations in snapping-shrimp dominated ambient noise

Passive acoustics is an effective method for monitoring marine mammals, facilitating both detection and population estimation. In warm tropical waters, this technique encounters challenges due to the high persistent level of ambient impulsive noise originating from the snapping shrimp present throughout this region. This study presents the development and application of a neural-network based detector for marine-mammal vocalizations in long term acoustic data recorded by us at ten locations in Singapore waters. The detector’s performance is observed to be impeded by the high shrimp noise activity. To counteract this, we investigate several techniques to improve detection capabilities in shrimp noise including the use of simple nonlinear denoisers and a machine-learning based denoiser to suppress the noise level. These are shown to enhance the detection performance significantly. Using the robust detectors developed, we discuss some of the vocalizations detected over three years of our acoustic recorder deployments.