Background Noise Research Articles

Improvement method for sound quality of pharyngeal speech by using Bayes' theorem

In general, a speech signal can be measured by a microphone and a pharyngeal microphone etc.. However, speech signal measured by a microphone often contains the background noise. On the other hand, though a pharyngeal microphone is effective for background noise, it involves body conducted noise. In this study, we propose an improvement method for sound quality of speech signal measured trough a pharyngeal microphone to achieve well the speech recognition. The relationship between the original speech signal and the speech measured by the pharyngeal microphone is not clear. Therefore, we consider the relationship as multiplicative and additive models of the original speech signal and noise components with unknown parameters. An algorithm to estimate simultaneously the original speech signal and the unknown parameters is proposed by using Bayes' theorem based on the measured speech through the pharyngeal microphone. Finally, a speech recognition experiment is conducted to confirm the effectiveness of the proposed algorithm.

Comparing Kurtosis-adjusted weighted levels with other metrics to assess the risk of hearing loss from non-Gaussian noise exposures

Several studies have shown that impulsive noise can cause more damage to hearing than steady-state noise of equal energy. As a result, a large body of research has been devoted to evaluating the hazards of impulsive noise. However, work environments often have varying noise patterns, including Gaussian background noise combined with high-level transient noises contributing to the worker's daily dose. Thus, an energy-based noise metric underestimates the risk of hearing loss unless incorporating a temporal structure correction term. Kurtosis has been reported as an effective adjunct to energy for predicting hearing hazards from non-Gaussian noise exposures. In this work, recordings of real impulsive noise were conducted using a portable measuring system at a very high sampling frequency and using high-pressure microphones. Controlled synthetic non-Gaussian noise was generated using these impulsive signals and steady-state background noise. After exploring the detailed temporal structure of the noise, the daily exposure time has been determined using different metrics, including kurtosis-adjusted weighted levels, weighted peak levels, and auditory risk units. The examples presented here aim to demonstrate how kurtosis can help develop more accurate methods to prevent noise-induced hearing loss.

Influence of the fundamental frequency and number of harmonics on preference judgements for fan sounds

Fan sounds are often heard in everyday life. Besides broadband noise, the sounds often contain one single or multiple tonal components, which can render the sound rather unpleasant and annoying. In this study, level adjustments for equal preference were measured in listening experiments for synthetic fan sounds compared to a fixed reference sound. The stimuli were based on two different background noises with one embedded single tone, or a multitone with eight partials. The tonal components had fundamental frequencies of either 80 or 240 Hz. All sounds containing tones needed level reduction to become equally preferred as the reference sound. The tonal configuration, as well as its fundamental frequency, did typically influence the effect. Additional level reductions for signals containing higher harmonics were necessary to reach equal preference compared to signals that contained only a single tone. For a fundamental frequency of 80 Hz, the additional level reductions were only 1.5 dB on average, for a fundamental frequency of 240 Hz up to 9.7 dB were necessary. The level reductions for equal preference were larger for sounds using a background averaged across real fan sounds compared to using bandlimited synthetic noise.

Development of operational tire-suspension contact force analysis using frequency based substructuring

With increase of EVs, we need to improve road noise performance as background noise caused by engines are reduced. On the other hand, to predict road noise performance, it is necessary to consider both tire and vehicle vibration characteristics, and various prediction techniques have been reported. Frequency based substructuring method (FBS) can predict the road noise performance by combining vibration characteristics of components of a vehicle, and is effective for reducing the development cost because it can clarify the development goals of both OEM and suppliers. With this regard, FBS is very useful for implementing MBD in the development of vibration noise. In this paper, for the road noise up to 500 Hz, in which structure-borne noise is the main source, we calculate the coupled vibration characteristics consisting of two subsystems: a tire (including a rigid wheel) and a suspension, by calculating in the rolling state for each subsystem by simulation and combined them by FBS. To evaluate the method, we compared the results with the vibration characteristics of the entire system in the rolling state. In particular, we calculated the transfer function of a tire in a rolling state to take the effect of tire rolling into account .

ADAPTIVE METHOD OF FORMING COMPLEX SIGNALS ENSEMBLES BASED ON MULTI-LEVEL RECURRENT TIME-FREQUENCY SEGMENT MODELING

The article investigates the implementation of an adaptive method for forming ensembles of complex signals based on multilevel recurrent time-frequency segmentation. It addresses the key challenges faced by cognitive wireless networks operating in dynamic radio frequency environments with high levels of interference, necessitating rapid adaptation to changes in the spectral characteristics of signals. The study substantiates the need for adaptive filters and specific transformations to enhance signal processing quality, particularly in environments with high variability in frequency characteristics and significant noise interference. The proposed method of multilevel recurrent time-frequency segmentation allows for the modification of time segment durations and the use of segments of varying lengths, providing flexibility in signal processing and adaptation to current conditions. This adaptability enables optimal signal processing for each individual case, taking into account short-term impulses, long-term fluctuations, and various types of noise and distortions. This approach effectively separates frequency components and reduces interference between them, which is crucial for maintaining high signal quality and communication stability in cognitive networks. It has been proven that the use of adaptive filters such as LMS (Least Mean Squares) and RLS (Recursive Least Squares), as well as fast Fourier Transform (STFT), wavelet, and Hilbert transforms at different stages of multilevel time-frequency segmentation, significantly enhances signal interference resistance and energy efficiency. Comparative analysis of signal metrics before and after filtering and transformation shows an increase in signal quality by 14,3–24,5% and a reduction in noise levels by 21,7–29,6%. The wavelet transform, in particular, proved to be highly effective, allowing for precise extraction of useful frequency components from the noise background and improving signal parameters through dynamic adjustment to specific radio environment conditions. Experimental results confirm the effectiveness of the proposed method, demonstrating its ability to ensure consistently high-quality processing of complex signal ensembles even in dynamic cognitive radio environments.

Aircraft icing detection via flow-induced ambient random noise: from theory to wind tunnel experiments

Aircraft icing is a serious threat to flight safety. This paper presents a passive approach to detect ice accreted on an airfoil using flow-induced random noise considering spatially-inhomogeneous transient excitations. Under the assumption of diffuse-field, by computing the cross-correlation of ambient vibration measured at two receivers and band-pass filtering it, the guided wave dispersion curve of the structure can be well-reconstructed, by which one can estimate the ice thickness by matching the reconstructed dispersion curve with its theoretical model. In practice, however, the diffuse-field assumption may not hold as spatially-inhomogeneous strong transient excitations exist. The associated directional wave propagation generates additional peaks in the cross-correlation and this contaminates the ice detection result. Therefore, this paper proposes a signal processing technique to identify and suppress the contributions from the transient source in the measurements, by which a reliable dispersion curve reconstruction and an accurate ice detection can be retrieved. The proposed method is shown to be efficient via a wind tunnel experiment. The passive ice detection method is specifically suitable for aircraft online monitoring because only passive sensors are needed which can be tiny, light, and mounted on the internal surface of fuselage skin without any influence on aircraft aerodynamics.

Assessing indoor and outdoor noise quality and associated health impacts at selected schools in Surat, India

An optimal ambient environment is crucial for healthy development, yet the excess of vehicles and inadequately planned educational infrastructures compromise cognitive growth in children. This study evaluates the impact of ambient noise on student cognition in five schools situated along city arterial and sub-arterial roads. A total of 717 students from grades 7, 8, 9, and 11 participated. Noise levels were measured in prevailing noisy conditions (PC) and controlled silence conditions (SC) within classrooms using the Kimo dB300 sound level meter. The observed differences in noise levels between PC and SC varied across grades, showing reductions of 4.7 ± 3.5 dB, 6.1 ± 0.7 dB, 6.7 ± 1.3 dB, and 4.4 ± 3.46 dB for grades 7, 8, 9, and 11, respectively. Statistical analyses revealed significant differences in the average noise levels under both conditions. Furthermore, cognitive task performance in PC and SC was significantly different, as evidenced by paired t-tests in tasks such as multiplication, word formation, reasoning, and sentence construction. Dimension reduction techniques identified room acoustics, task difficulty, speech intelligibility, mental deficiency, and annoyance as critical factors affecting learning abilities. Teachers reported that major impacts of noise include annoyance, sleeplessness, headache, and hearing impairment.

The impact of marine traffic radiated underwater noise on the in the Baltic Sea

Continuous anthropogenic noise, generated by commercial shipping, is considered to exert a considerable pressure on the marine environment. Noise emitted by ships is mainly due to their propellers whose energy is predominantly at low frequencies and can interfere with the communication of marine animals. The EU Marine Strategy Framework Directive (MSFD), has set the objective of achieving good environmental status of Europe's seas. In particular, anthropogenic underwater sound should be at a level that does not harm the marine environment. Here we present the first implementation of the new framework for assessing the environmental status of low frequency continuous noise on a basin scale, carried out in the frames of the EU-funded HELCOM BLUES project. The assessment was based on modeled levels in two 125 and 500 Hz decidecade bands deemed relevant for the indicator fish and marine mammals respectively. Sound propagation modelling, calibrated against underwater ambient noise monitoring data, results in statistical sound maps in species-specific frequency bands. The application of spatial thresholds allows the environmental status of marine areas to be determined. Factors affecting the reliability of results are discussed.

Automatic estimation of the wind turbine noise with recurrent neural networks

There is growing interest in the development of renewable energies, particularly wind power. However, wind turbines generate noise that can affect the sound environment of nearby residents. This study focuses on the isolation of wind farm noise (WFN) from the surrounding ambient noise. Our method is based on a Recurrent Neural Network (RNN) Architecture that captures temporal dependencies in the acoustic signal. This proposal is compared to Non-Negative Matrix Factorization (NMF) that has shown first promising results on a previous study on simulated sound scenes. Our approach relies on Long Short Term Memory (LSTM) RNN, conducted using an end-to-end trained model and a Gated Recurrent Unit (GRU). The training and testing dataset is constructed by superimposing measured background noise with synthesized wind turbine noise, effectively simulating realistic environmental conditions. This study aims to reduce the acoustic impact of wind turbines on communities and attempt to control turbine operation using advanced machine learning techniques.

Ambient noise criteria framework for future environmental noise mitigation targets in Singapore

This study aims to establish a framework for outlining future noise mitigation strategies tailored to the unique context of Singapore and evaluates potential precautions. Singapore, as a city-state, is known as a 'city in a garden,' situated in the typically tropical climate. By incorporating a comparative analysis of noise criteria from local guidelines across countries from Europe, America, and Asia, the study aims to place Singapore's acoustic limits within an international context. This comparison, enhanced with the World Health Organization's health and well-being standards, seeks to establish a comprehensive framework for acoustical design assessment in residential, office, and educational buildings targeting resting and cognitive functions. In conclusion, starting from a global perspective, the study points to the importance of focusing on the local context and acknowledging the challenges and benefits that the context brings for acoustical comfort. The study discusses the importance of context and content in crafting acoustical design criteria, considering the impact of biophilic design and active noise cancellation windows to achieve adaptive acoustic comfort in Singapore's built environment.

Dual-stream Noise and Speech Information Perception based Speech Enhancement

In real-world scenarios, dynamic ambient noise often degrades speech quality, highlighting the need for advanced speech enhancement techniques. Traditional methods, which rely on static embeddings as auxiliary features, struggle to address the complexities of varying noise conditions. To overcome this, we propose a Dual-stream Noise and Speech Information Perception (DNSIP) approach that dynamically detects and processes both noise and speech through innovative information extraction and suppression mechanisms. Initially, non-speech segments predominantly contain environmental noise, while speech segments carry information about the intended speaker. To handle this dynamic nature, real-time voice activity detection (VAD) is employed to accurately differentiate between speech and noise components. Building on VAD estimates, we propose an innovative information extraction framework that selectively extracts relevant noise and speech features from the noisy input, establishing a dual-stream network for concurrent noise and speech learning. To account for the temporal and spectral variability of noise and speech, a frequency-sequence attention mechanism is integrated, enhancing the model’s ability to learn contextual and spectral dependencies. Additionally, an information suppression module is introduced to minimize cross-stream interference by attenuating noise within the speech stream and suppressing speech content within the noise stream. The derived noise and speech spectrograms are then utilized to formulate a minimum mean square error log-spectral amplitude (MMSE-LSA) estimator for robust speech enhancement. Experimental evaluations on the WSJ0 and VCTK+DEMAND datasets demonstrate that our DNSIP approach surpasses existing state-of-the-art methods, underscoring its efficacy in challenging acoustic environments.

Fundamental study on ambient noise reduction for analysis of hitting sound of professional baseball players

In recent years, the sound of a hitting ball during a baseball game has focused attention on the American Major League Baseball and Japanese professional baseball. It is reported that information such as the direction and distance of a batted ball can be detected from the sound. The sounds of players hitting the ball are sometimes expressed using various onomatopoeic. In our previous study, we recorded the sound of a baseball hit during hitting practice for players of a university baseball team in the Tokyo metropolitan area in Japan. Relationships between the acoustic characteristics of the sound and the distance and direction of the ball were clarified. In this research, acoustic analysis of professional baseball players' batted ball sounds from Youtube and other videos is used to identify the acoustic features and characteristics of the batted ball. This paper describes the method and accuracy of noise removal around the hitting sound extracted from the video using a convolutional neural network with images.

Dynamic localization of unwanted events with machine learning-based method

In environmental acoustics measurements it is necessary to separate the contribution of the investigated source from the residual noise, by means of ambient noise measurements. Source separation is a complex task, especially for wind turbine noise, where long-term measurement campaigns are required. Although data analysis procedures that provide for this task are already in use, all of them require a preliminary step for anomalous events removal, i.e. events which are not characteristic of standard residual noise. This step can be complex and very time-expensive depending on the duration of the measurements and the measurement site. Depending on the specific site and source under evaluation, the variety of sounds that can be considered as unwanted for compliance with environmental acoustic regulations is heterogeneous, ranging from road vehicles, trains and planes to human and animal sounds and more. Nowadays, it is possible to extrapolate from an audio source the features that describe its semantics using machine learning. The approach presented in this paper aims to localize in time a specific event within a measurement containing several acoustic events, in order to label it and then proceed to remove the spurious events from the measurement.

The Correlation analysis between ambient noise level at low-frequency range and wind speed in shallow water

The ambient noise level in the ocean is an important consideration in the enhancement of passive sonar detection performance as well as undersea surveillance in a harbor area. The wind affects considerably the ambient noise level in the ocean. For this reason, many researchers have analyzed the wind dependency of ambient noise levels in the ocean in the last few decades. However, this analysis is mainly limited to more than the frequency of 100Hz, because the ambient noise level below 100 Hz has been affected by sea current. In this study, we have investigated ambient noise levels in the frequency range 10 Hz - 100Hz as the wind speed increases in shallow water. A drifting small vessel is used to exclude the effect of fluid noise generated on the surface of a hydrophone due to the sea current. The correlation between the ambient noise level and wind speed in frequency range 10 - 25 Hz showed from 0.77 to 0.95. Whereas, it was from 0.11 to 0.69 in the frequency range 32 -100 Hz. The maximum correlation appeared at 10 and 16 Hz. This was much higher than those measured in the frequency range 1kHz - 25 kHz.

Observation of the modal propagation of T waves using distributed acoustic sensing

T waves are low-frequency acoustic-waves (< 100 Hz) generated by earthquakes and propagating in the watercolumn. It is largely accepted that T waves propagate as modes. However, the direct observations of this modal propagation in experimental data remain surprisingly rare. Observation of T waves using distributed acoustic sensing (DAS) makes possible space-sampling with one sensor every few meters over tens of kilometers. Using a simple 2D Fourier transform, the time-space pressure-field can be turned into a frequency-wavenumber field where the propagating wave dispersion curves are revealed. Using freely accessible data from a DAS experiment on the coast of Oregon (USA), we show that the dispersion curves of T waves are consistent with the presence of four propagating modes. The number of modes decreases as expected as T waves propagate shoreward with decreasing depths. The T-wave mode dispersion curves can be fitted by using environmental parameters obtained through ambient noise Scholte wave inversion (also visible in the data).

DOL3: Distilled OpenL3 audio embeddings for lightweight audio classification

Deep audio representations, also known as embeddings, recently became a popular alternative to conventional features like spectrograms for a wide range of audio classification tasks because of their domain-agnostic character and reduced training costs. Still, the usage is often limited to rather computationally intensive system due to the nature of their extraction from large networks. This paper aims to minimize the computational costs of embedding extraction by distilling the knowledge of the OpenL3 audio network to a smaller student network. Results show that the student network maintains comparable performance as the teacher network on various music and ambient noise classification tasks, while reducing the network size by over 90\% and the computational load by five times.

Long term wind turbine sound monitoring

In the DLR Windpark Wiwaldi, consisting of two wind turbines, extensive measurements of sound and meteorology have been conducted. In the context of long-term wind turbine sound monitoring, the data from such measurements are used to understand and decipher physical effects. This kind of data suffers sometimes for reliability, because of many changing conditions and situations around the microphone. Therefore, in three different phases the sound of various sources were captured. Before the turbines were installed, initial measurements were made to identify background noise, followed by additional measurements to assess the impact of installing weather masts on the sound field. Ultimately, the wind turbines were operated reporting the exact state of the turbines (such as rotation speed, pitch, yaw, etc.) to relate to the associated sound levels. This presentation illustrates how all these measurements are used to create a consistent picture of the noise situation around the wind park depending on specific weather situation and discusses the need of accuracy. We document the limitations and uncertainties associated with the measurements and show these limitations when interpreting and reporting the results to ensure users in understanding the reliability and confidence level of such data.

Noise reduction design of new residential buildings along urban roads-taking a residential district in Foshan as an example

Taking a residential district in Foshan along the city road as an example, the impact of urban road traffic noise on it is studied. Firstly, the data is monitored in the noise field, then three-dimensional simulation models of residential buildings and road distribution are established in Cadna/A software, and noise simulation prediction is carried out according to the actual situation and relevant norms and standards. The predicted noise values of the exterior surface of residential buildings are obtained, which has exceeded the permissible interior noise level for residential buildings. By comparing the application effects of noise reduction measures in buildings, the noise reduction design scheme of new residential buildings along urban roads is finally proposed: the sound insulation doors and windows that meet the requirements are adopted. After the completion of the house, the indoor background noise and the sound insulation performance of the doors and windows are measured. The measured results show that the selection of sound insulation doors and windows is suitable and the indoor noise environment meets the design requirements of noise reduction. The results are helpful to provide technical reference for the noise reduction design and renovation of other residential buildings along urban roads.

Understanding human responses to air source heat pump noise: examining tonal changes and operating cycles

Air Source Heat Pumps are pivotal in the Government of the United Kingdom's target of achieving net zero carbon emissions by 2050. However, their widespread adoption faces challenges, with noise being among the main barriers to entry and a restricting factor at the planning stages. This study addresses concerns regarding noise propagation from outdoor units to indoor environments through a two-part experiment, aiming to capture human responses to Air Source Heat Pump noise based on the changes in the characteristics of noise emissions at various operating conditions, background noise levels, and source distances. The first part evaluates peoples' responses under static operating conditions to 20-second recordings, using the Self-Assessment Manikin scale and an annoyance scale. Following each recording, they used the Self-Assessment Manikin pictorial scale and an annoyance scale to report their response. In the second part, responses to transient Air Source Heat Pump operation cycles are assessed using 60-second recordings. Participants evaluated each recording at two points: after a transition in the operating cycle and at the end of the recording. The findings of the study provide insights into the human response to heat pump noise, essential for developing strategies to facilitate wider Air Source Heat Pump adoption.

Room acoustic survey of open and enclosed learning spaces in Finland

Use of open learning spaces for education has increased in schools while enclosed learning spaces are still used in most schools in overall. There is some debate that these two learning space types might differ in acoustic perceptions. If they do, room acoustic quality might be one explaining factor. Therefore, our purpose was to determine, whether these learning space types differ in room acoustic properties. We studied 10 schools with mostly enclosed and another 10 schools with open or flexible learning spaces in Finland. Altogether 73 learning spaces were investigated among the 20 schools. Measurements concerned reverberation time, Speech Transmission Index, and background noise level. Most of enclosed learning spaces (63%) met the new requirements of Finnish building regulations for enclosed learning spaces. Opposite to that, most of open learning spaces (91%) did not meet the regulations for open learning spaces. The results can be used in the root cause analysis of a teacher survey conducted in the same 20 schools. Teacher survey will be reported in another study. Parallel analysis of objective and subjective results is necessary to judge whether the current regulations are justified.