Acoustic Events Research Articles

Detection of Water on Road Surface with Acoustic Vector Sensor.

This paper presents a new approach to detecting the presence of water on a road surface, employing an acoustic vector sensor. The proposed method is based on sound intensity analysis in the frequency domain. Acoustic events, representing road vehicles, are detected in the sound intensity signals. The direction of the incoming sound is calculated for the individual spectral components of the intensity signal, and the components not originating from the observed road section are discarded. Next, an estimate of the road surface state is calculated from the sound intensity spectrum, and the wet surface detection is performed by comparing the estimate with a threshold. The proposed method was evaluated using sound recordings made in a real-world scenario, and the algorithm results were compared with data from a reference device. The proposed algorithm achieved 89% precision, recall and F1 score, and it outperforms the traditional approach based on sound pressure analysis. The test results confirm that the proposed method may be used for the detection of water on the road surface with acoustic sensors as an element of a smart city monitoring system.

Deep Learning-based drone acoustic event detection system for microphone arrays

Deep Learning-based drone acoustic event detection system for microphone arrays

Эмоциональная составляющая воспринимаемого качества акустических событий

<p>The article represents the description of the results of the empirical research of the peculiarities of the acoustic events interpretation by the tuvans living in different regions of the Republic of Tuva. This article includes the analytical statistics of the emotions that have been felt by the respondents while listening to the natural sounds of the surrounding area. It has been revealed that the research participants prefer describing the perceived acoustic events in terms of “pleasant-unpleasant” and the most widely spread basic emotions are fear and happiness.</p>

A Note on Performance Assessment of Signal Energy–Based Acoustic Source Localization in a Carbon Fiber–Reinforced Polymer Plate

Abstract The effectiveness of the signal energy–based acoustic source localization approach in practical applications has yet to be established. This is addressed herein by conducting an experimental study on a 500 mm × 500 mm carbon fiber–reinforced polymer plate and generating artificial acoustic events in the plate. Upon acquiring the propagating wave signals at several well-scattered sensors, the signal energy–based approach is applied, and the accuracy of the source localization results is noted. Seven experiments are performed with varying source locations, sensor-plate bonding, and excitation types. This approach has performed well for five experiments with source localization errors below 15 mm. The remaining two experiments where the acoustic sources are relatively close to the plate edges compared to the other experiments have, however, produced large localization errors, indicating a scope of improvement in the approach to encompass all situations.

ANALYSIS OF THE SOUNDSCAPE OF SKI RESORTS IN BAD HOFGASTEIN (AUSTRIA) AND BIAŁKA TATRZAŃSKA (POLAND)

Aim of the study: Firstly, due to the different sizes and spatial organization of ski resorts, it is important to determine whether there is a local differentiation of the sound landscape within one resort. Secondly, due to the continuous expansion of the ski infrastructure, it is essential to develop methods for improving the quality of the acoustic environment. It is therefore necessary to determine whether there are acoustic differences between the different ski resorts. Material and methods: The research we conducted consisted in measuring the sound levels, and inventorying various types of acoustic events. The examined ski resorts, despite the difference in size of the areas they occupy and the numbers of users, are in many aspects comparable to each other. Results and conclusions: Ski resorts occupying a relatively small area but with a high concentration of infrastructure and skiers may exert a stronger environmental impact than resorts with a larger area and lesser density of infrastructure as well as a lower number of users per area. The skiers themselves, their manners, and their environmental consciousness have a great impact on the quality of the sound landscape.

Matching visual and acoustic events to estimate detection probability for small cetaceans in the ACCOBAMS Survey Initiative

Estimating the detection probability of small cetaceans using either visual or acoustic surveys is difficult because they do not surface or vocalise continuously and can be imperceptible to an observer or hydrophone. Animals seen at the surface may have lower vocalisation rates, while submerged individuals may be more vocally active. This study aims to estimate visual, acoustic and combined detection probability by using Mark-Recapture Distance Sampling (MRDS) methodology. We used vessel-based visual sightings and acoustic data (based on click identification) collected simultaneously during the ACCOBAMS Survey Initiative in summer 2018 onboard theR/V Song of the Whale. This study focused on small cetaceans in the Mediterranean Sea, including the most commonly-encountered species, the striped dolphin(Stenella coeruleoalba). We identified duplicate events between visual and acoustic platforms using a decision tree based on time and distance thresholds to estimateg(0) (the detection probability on the trackline) for small cetaceans. A total of 30 duplicate events were identified from 107 and 109 events identified by the visual and acoustic platforms respectively. We tested the models with two key functions. With a hazard-rate key function, theg(0) was estimated at 0.52 (CV=21.0%) for both platforms combined, 0.29 (CV=25.6%) for the visual platform and 0.32 (CV=25.1%) for the acoustic platform. With a half-normal key function,g(0) was estimated at 0.51 (CV = 21.7%) for both platforms combined, 0.29 (CV = 25.6%) for the visual platform and 0.33 (CV = 23.2%) for the acoustic platform. Our results illustrate that passive acoustic monitoring can be used as an independent platform in MRDS to estimate the detection probability. Our estimate ofg(0) was well below 1, far from the perfect detection commonly assumed for abundance estimation. Without correction for detection biases, total abundance would be underestimated by a factor of two when using both acoustic and visual data. This highlights the importance of using dual-platform surveys to estimate detection probability in order to improve abundance estimates and conservation efforts.

Experimenting ‘In Vivo’ with Dynamic Glottal Parameters: Methodological Issues, Technical Tips & Tricks and Preventable Pitfalls

‘In vivo’ studies pertaining to dynamics of vocal fold vibration motion, to vocal fold contact and collision, to vocal onset and offset and to mechanical efficiency all need valid, sensitive and precise measurements of the different mechanical parameters involved. This is also true for investigating the physiological correlates of particular acoustic events like register breaks or diplophonia. The main physical parameters involved are: vocal fold movement and shaping, particularly the velocity of tissue displacement, glottal area, tissue distortion, intraglottal pressure, transglottal air flow, vocal fold contact and collision stress, etc. This article presents a critical review of the instruments and techniques involved in the direct measurements of the glottal dimensions and movements, the transglottal airflow, the VF contact changes, the pressures and the sound acoustic pressure. In each case are analyzed the methodological aspects that are critical for validly calibrating and synchronizing these signals, and correcting them for time delays. Moreover, it is shown how new parameters, like vocal fold velocity, intraglottal pressure, vocal fold collision stress, can be inferred from these (raw or after differentiation) signals by combining them. Finally, the discussion focuses on weighing advantages and limitations of techniques for monitoring the glottal area, i.e. photometry and the high-speed imaging, the latter involving the relevance in this scope of future developments in endoscopic and external imaging techniques, and in image processing software. Our aim is to facilitate the work of future researchers by showing how to solve important technical pitfalls, how to apply the necessary corrective measures - and which ones - where needed, and how to get the most out of combinatorial measures.

Greedy regression and differential convex-based deep learning for audio event classification

Audio Event Detection (AED) and classification of acoustic events has become a notable task for machines to interpret the auditory information around us. Nevertheless, it has been a difficult and cumbersome task to extract the most basic characteristics of acoustic events that encapsulate the fundamental elements of the audio events. Previous works on audio event classification utilized supervised pre-training as well as meta-learning approaches that happened to depend on labeled data therefore facing instability. Deep Learning is progressing in an increasingly mature direction, and the application of deep learning methods to detect acoustic event has become more and more sought after. The proposed hybrid method called Greedy Regression-based Convolutional Neural Network and Differential Convex Bidirectional Gated Recurrent Unit (GRCNN-DCBGRU) is introduced to learn a vector representation of an audio sequence for Audio Event Classification (AEC). Differential Convex Bidirectional Gated Recurrent Unit is analogous to long short-term memory and involves time-cyclic long-term dependencies with a lesser processing complexity. The model first extracts acoustic features from the sound event dataset through a Differential Convex Bidirectional Gated Recurrent Unit employing Gabor Filter bank features and then extracts the local static acoustic features through the Greedy Regression-based Convolutional Neural Network by utilizing Mel Frequency Cepstral Coefficients (MFCC). Finally, the Differential Convex Meta-Learning classifier is used for the final acoustic event classification. Extensive evaluation on large-size publicly available acoustic event database like Findsounds2016 will be performed in Python programming language to demonstrate the efficiency of the proposed method for the AEC task. To demonstrate the visualizations of individual modules and their influence on overall representation learning for AEC tasks, several parameters like audio detection time, audio detection accuracy, precision, and recall are measured.

Research on Precursor Information of Brittle Rock Failure through Acoustic Emission

Dynamic failure of surrounding rock often causes many casualties and financial losses. Predicting the precursory characteristics of rock failure is of great significance in preventing and controlling the dynamic failure of surrounding rock. In this paper, a triaxial test of granite is carried out, and the acoustic emission events are monitored during the test. The fractal characteristics of acoustic emission events’ energy distribution and time sequence are analyzed. The correlation dimension and the b value are used to study the size distribution and sequential characteristics. Furthermore, a rock failure prediction method is proposed. The correlation dimension is chosen as the main index and the b value is chosen as a secondary index for the precursor of granite failure. The study shows that: (1) The failure process can be divided into an initial stage, active stage, quiet stage, and failure stage. (2) The b value and correlation dimension both can describe the process of rock failure. There is a continuous decline before failure. Because of the complexity of the field, it is difficult to accurately estimate the stability of surrounding rock using a single index. (3) The combination of the b value and correlation dimension to establish a new method, which can accurately represent the stability of the surrounding rock. When the correlation dimension is increasing, the surrounding rock is stable with stress adjusting. When the correlation dimension is decreasing and the b value remains unchanged after briefly rising, the surrounding rock is stable, and stress is finished adjusting. When the correlation dimension and b value are both decreasing, the surrounding rock will be destroyed.

Modeling reduced speech using a simulated vocal tract and overlapping acoustic events

Reduced utterances are characterized by the alteration or deletion of segments (i.e., syllables, words) and often occur in spontaneous or casual speech styles. These reductions are common in everyday communication and are generally understood given sufficient context. However, modeling reduced speech is not trivial, as the effects of context, articulatory limitations, idiosyncrasies, and speech rate need to be considered in order to approximate acoustic observations of reduction. We utilize a speech production model which modulates a simulated vocal tract according to acoustic events organized along a time axis. The present study focuses on validating its capacity for modeling reduced speech. Evidence of its effectiveness is provided by way of synthetic example utterances, each accompanied by a description of the steps taken to synthesize the utterance and acoustic comparison to an unreduced equivalent. Word-medial consonant reductions will be studied first before proceeding to massive reductions where entire syllables and words are affected. We iteratively synthesize reduced utterances until they are evaluated through informal listening to be comparable to real acoustic observations. The results of this study are discussed as validation of the idea that acoustic events can be arranged to model reductions.

Passive ocean acoustic waveguide remote sensing of vocalization behavior and spatial distribution of diverse marine mammal species in the Norwegian and Barents Sea

Leveraging data acquired using a 160-element coherent hydrophone array deployed in the Norwegian and Barents Seas during spring 2014, and the passive ocean acoustic waveguide remote sensing (POAWRS) technique is employed to enable instantaneous wide-area monitoring of marine mammal vocalizations over expanses exceeding 100 km in diameter. The vocalization behavior of diverse marine mammal species including Fin, Humpback, Minke, Sperm, and Beluga whales are analyzed, quantifying time-frequency characteristics and call patterns from their vocalization signals present in high-resolution beamformed power spectrograms. Previously developed automatic detection and machine learning algorithms are employed for clustering and classification of underwater acoustic events, facilitating the subsequent manual audio-visual inspection of whale calls. Bearing-time trajectories of repetitive species-specific vocalizations signals are utilized to estimate locations of the whale calls and hence their temporo-spatial distributions. Through comparative studies, we assess the presence and abundance of marine mammal species-specific vocalization signals in three distinct coastal regions off Norway, namely, Alesund, Lofoten, and Northern Finmark. In addition, we investigate correspondences of marine mammal distributions with concurrently imaged instantaneous wide-area fish distributions from distinct fish species to elucidate potential predator-prey interactions and habitat preferences.

Human auditory ecology: Extending hearing research to the perception of natural soundscapes by humans

A “natural soundscape” refers to the case where the contribution of acoustic events resulting from human activity can be considered as negligeable. As a consequence, natural soundscapes are only composed of biological sounds and geophysical sounds shaped by the specific way sounds propagate within the habitat under study. Within this framework, studying soundscape perception in humans aims at unveiling the relationship between the features of sound mixtures picked up at a given place and time by the peripheral auditory system of a human listener and the characteristics of the auditory percept evoked by these features. We will present a research program based on large and ecologically-valid acoustic databases recorded in protected areas aiming to (i) better understand the mechanisms involved in auditory perception of natural soundscapes; (ii) characterize and explain the effects of sensorineural hearing loss on perception of natural soundscapes; and (iii) assess the extent to which alterations in soundscape perception can be restored back to normal via hearing aids. This programme combines modelling and psychophysical methods to explore our ability to distinguish between habitat, time of day and season, to detect the presence of biological sound sources and to assess levels of biodiversity in the habitat.

Spatio-temporal evolution of acoustic emission events and initiation of stress fields in the fracturing of rock mass around a roadway under cyclic high-stress loading.

To study fracture mechanisms and initiation of stress fields in the rock mass around a roadway subjected to cyclic stress, a series of loading and unloading tests were conducted on the rock mass around the roadway by using high-precision acoustic emission (AE) monitoring. The results show that intense AE activities occur in a specimen during cyclic load-holding at different levels. With the increase in the number of cycles, the overall stability of the specimen gradually decreases. In the cyclic loading and unloading process, the specimen exhibits a Kaiser effect. As the number of cycles increases, more AE events occur in the unloading stage and a Felicity effect is manifest. The spatial distribution of AE events is related to the stress regime and structure of the specimen, crack propagation in the roadway exhibits directionality due to effects of the principal stress. High stress is conducive to microcrack initiation and propagation in the specimen, which accelerates damage accumulation and macrofracture formation in a rock mass. The research provides a reference for roadway support work and disaster prevention and control in deep mines.

Experimental study on the interrelation of multiple mechanical parameters in overburden rock caving process during coal mining in longwall panel

In order to comprehend the dynamic disaster mechanism induced by overburden rock caving during the advancement of a coal mining face, a physical simulation model is constructed basing on the geological condition of the 21221 mining face at Qianqiu coal mine in Henan Province, China. This study established, a comprehensive monitoring system to investigate the interrelations and evolutionary characteristics among multiple mechanical parameters, including mining-induced stress, displacement, temperature, and acoustic emission events during overburden rock caving. It is suggested that, despite the uniformity of the overburden rock caving interval, the main characteristic of overburden rock lies in its uneven caving strength. The mining-induced stress exhibits a reasonable interrelation with the displacement, temperature, and acoustic emission events of the rock strata. With the advancement of the coal seam, the mining-induced stress undergoes four successive stages: gentle stability, gradual accumulation, high-level mutation, and a return to stability. The variations in other mechanical parameters does not synchronize with the significant changes in mining-induced stress. Before the collapse of overburden rock occurs, rock strata temperature increment decreases and the acoustic emission ringing counts surges with the increase of rock strata displacement and mining-induced stress. Therefore, the collaborative characteristics of mining-induced stress, displacement, temperature, and acoustic emission ringing counts can be identified as the precursor information or overburden rock caving. These results are in good consistent with on-site situation in the coal mine.

Sound Event Detection with Perturbed Residual Recurrent Neural Network

Sound event detection (SED) is of great practical and research significance owing to its wide range of applications. However, due to the heavy reliance on dataset size for task performance, there is often a severe lack of data in real-world scenarios. In this study, an improved mean teacher model is utilized to carry out semi-supervised SED, and a perturbed residual recurrent neural network (P-RRNN) is proposed as the SED network. The residual structure is employed to alleviate the problem of network degradation, and pre-training the improved model on the ImageNet dataset enables it to learn information that is beneficial for event detection, thus improving the performance of SED. In the post-processing stage, a customized median filter group with a specific window length is designed to effectively smooth each type of event and minimize the impact of background noise on detection accuracy. Experimental results conducted on the publicly available Detection and Classification of Acoustic Scenes and Events 2019 Task 4 dataset demonstrate that the P-RRNN used for SED in this study can effectively enhance the detection capability of the model. The detection system achieves a Macro Event-based F1 score of 38.8% on the validation set and 40.5% on the evaluation set, indicating that the proposed method can adapt to complex and dynamic SED scenarios.

Polyphonic Sound Event Detection Using Mel-Pseudo Constant Q-Transform and Deep Neural Network

The task of identification of sound events in a particular surrounding is known as Sound Event Detection (SED) or Acoustic Event Detection (AED). The occurrence of sound events is unstructured and also displays wide variations in both temporal structure and frequency content. Sound events may be non-overlapped (monophonic) or overlapped (polyphonic) in nature. In real-time scenarios, polyphonic SED is most commonly seen as compared to monophonic SED. In this paper, a Mel-Pseudo Constant Q-Transform (MP-CQT) technique is introduced to perform polyphonic SED to effectively learn both monophonic and polyphonic sound events. A pseudo CQT technique is adapted to extract features from the audio files and their Mel spectrograms. The Mel-scale is believed to broadly simulate human perception system. The classifier used is a Convolutional Recurrent Neural Network (CRNN). Comparison of the performance of the proposed MP-CQT technique along with CRNN is presented and a considerable performance improvement is observed. The proposed method achieved an average error rate of 0.684 and average F1 score of 52.3%. The proposed approach is also analyzed for the robustness by adding an additional noise at different Signal to Noise Ratios (SNRs) to the audio files. The proposed method for SED task has displayed improved performance as compared to state-of-the-art SED systems. The introduction of new feature extraction technique has shown promising improvement in the performance of the polyphonic SED system.

An Experimental Investigation on the Foliation Strike-Angle Effect of Layered Hard Rock under Engineering Triaxial Stress Path.

Deep underground engineering encounters substantial layered hard rock formations, and the engineering triaxial stress path involves an increase in maximum principal stress, constant intermediate principal stress, and a decrease in minimum principal stress. However, previous research has focused on rock layer angles under conventional triaxial stress conditions, disregarding the influence of foliation strike angles in engineering triaxial stress scenarios. This study experimentally investigates the effects of foliation strike angles on layered hard rock under an engineering triaxial stress path. To account for the brittleness of layered hard rock, we propose a specific small sample-processing method tailored to the foliation strike angle. True triaxial loading tests are conducted on steep, thin slate samples with two different loading orientations, accompanied by acoustic emission monitoring. Results indicate that the strength under a traditional true triaxial compression condition is similar for specimens with 90° and 0° strike angles. Stress-strain curves show that larger deformations occur perpendicular to bedding planes, while surface fractures propagate exclusively along the bedding planes. Mechanical responses differ significantly between specimens subjected to the engineering triaxial stress path at 0° and 90° strike angles compared to conventional true triaxial loading tests, with a lower bearing capacity and differentiated intermediate and minimum principal strains in the 0° case. Conversely, the 90° case exhibits a higher bearing capacity, consistent deformation, and more acoustic emission events. Numerical simulations comparing plastic zone sizes during actual underground excavation support these conclusions. These findings highlight the effects of foliation strike angles, favoring the 90° strike-angle configuration for excavation activities and providing enhanced stability in the surrounding rock mass.

Self-supervised learning representation for abnormal acoustic event detection based on attentional contrastive learning

Most abnormal acoustic event detection (AAED) is completed by supervised training of deep learning methods, but manually labeled samples are costly and scarce. This work proposes a self-supervised learning representation for AAED based on contrastive learning to overcome the abovementioned problem. Auditory and visual data augmentations are applied simultaneously to create positive sample pairs. An attention mechanism is introduced into the encoder during self-supervised pre-training. A comparison between fused features by discriminant correlation analysis and a single feature is made to verify the ability of feature grasping for the self-supervised pre-trained model. The pre-training is completed on an abnormal acoustic dataset with noise. Research results show that the self-supervised pre-trained model can achieve an accuracy of 87.72% in linear evaluation and 88.70% in the downstream task with a pure small AAED dataset, which directly exceeds the results of supervised learning. This work releases the stress of the demand for abnormal acoustic event labels.

Characterization and cross-comparison of digital stethoscopes for telehealth remote patient auscultation

— Stethoscopes are commonly used to diagnose cardiac and respiratory diseases. The advent of digital stethoscopes added the capability to remotely monitor the cardiorespiratory well-being of patients and facilitate digital remote auscultations for physicians to examine their patients outside of the clinic. To increase the adoption of a hybrid diagnosis between in-person and remote auscultations, it is important to characterize the frequency response of different configurations of digital stethoscopes. A standard testing platform was adopted from the literature and re-validated by seven physicians. Two wearable and two non-wearable digital stethoscopes were investigated. A similarity factor score was used to analyze select time-synchronized acoustic events that achieved high fidelity between stethoscopes. The frequency responses between devices were compared using the correlation coefficient. Two devices reported the highest correlation coefficient of 0.72 and 0.75 in the wearable and non-wearable categories, respectively. The correlation coefficient of the frequency responses between the same pair of devices on the human chest was 0.91. With such a high agreement in the frequency response between the wearable and non-wearable digital stethoscope, it can be concluded that the configuration of digital stethoscope is not a limiting factor of performance, and both are acceptable and desirable for physicians to adopt either configuration in remote delivery of healthcare.

Bubble rupture & viability of red blood cells under resonant acoustic standing waves

Objective: The presentation of a novel prospective treatment for scenarios where bubble presence in the bloodstream poses a clinical risk. The method relies on generating resonant acoustic standing waves within a limb to non-invasively accelerate dissolution of bubbles present in the bloodstream via bubble rupture. Additionally, a preliminary assessment of the effects of the resonant acoustic waves and bubble rupture events on red blood cell viability is provided. Methods: Two semicircular piezoelectric (PZT) transducers electrically connected to each other were assembled around a small-girth segment of a rear thigh removed from a swine specimen. When driven at the frequency of electric resonance, this swine thigh and PZT transducer arrangement generates resonant acoustic standing waves within the swine thigh. Consequently, mechanical resonance of the system was non-invasively established by monitoring the electric response of the PZT to the applied frequency. The resonant acoustic field generated was used for the detection and rupture of bubbles that travel through a simulated blood vessel installed across the swine thigh. Two sets of experiments were carried out using this methodology, one with the artificial blood vessel filled with saline solution and one with defibrinated sheep blood. For the latter case, a preliminary hematologic assessment was done with red blood cell counts. Conclusion: Resonant acoustic standing waves effectively rupture bubbles of 300µm to 900µm within a simplified swine thigh model. The average dissolved gas content was 44% due to resonant acoustic waves at powers above 20W. No significant effect on red blood cell counts was observed.