Soundscape Characterization Using Autoencoders and Unsupervised Learning.

Passive acoustic monitoring (PAM) through acoustic recorder units (ARUs) shows promise in detecting early landscape changes linked to functional and structural patterns, including species richness, acoustic diversity, community interactions, and human-induced threats. However, current approaches primarily rely on supervised methods, which require prior knowledge of collected datasets. This reliance poses challenges due to the large volumes of ARU data. In this work, we propose a non-supervised framework using autoencoders to extract soundscape features. We applied this framework to a dataset from Colombian landscapes captured by 31 audiomoth recorders. Our method generates clusters based on autoencoder features and represents cluster information with prototype spectrograms using centroid features and the decoder part of the neural network. Our analysis provides valuable insights into the distribution and temporal patterns of various sound compositions within the study area. By utilizing autoencoders, we identify significant soundscape patterns characterized by recurring and intense sound types across multiple frequency ranges. This comprehensive understanding of the study area's soundscape allows us to pinpoint crucial sound sources and gain deeper insights into its acoustic environment. Our results encourage further exploration of unsupervised algorithms in soundscape analysis as a promising alternative path for understanding and monitoring environmental changes.

Pied tamarins change their vocal behavior in response to noise levels in the largest city in the Amazon.

Many animal species depend on sound to communicate with conspecifics. However, human-generated (anthropogenic) noise may mask acoustic signals and so disrupt behavior. Animals may use various strategies to circumvent this, including shifts in the timing of vocal activity and changes to the acoustic parameters of their calls. We tested whether pied tamarins (Saguinus bicolor) adjust their vocal behavior in response to city noise. We predicted that both the probability of occurrence and the number of long calls would increase in response to anthropogenic noise and that pied tamarins would temporally shift their vocal activity to avoid noisier periods. At a finer scale, we anticipated that the temporal parameters of tamarin calls (e.g., call duration and syllable repetition rate) would increase with noise amplitude. We collected information on the acoustic environment and the emission of long calls in nine wild pied tamarin groups in Manaus, Brazil. We found that the probability of long-call occurrence increased with higher levels of anthropogenic noise, though the number of long calls did not. The number of long calls was related to the time of day and the distance from home range borders-a proxy for the distance to neighboring groups. Neither long-call occurrence nor call rate was related to noise levels at different times of day. We found that pied tamarins decreased their syllable repetition rate in response to anthropogenic noise. Long calls are important for group cohesion and intergroup communication. Thus, it is possible that the tamarins emit one long call with lower syllable repetition, which might facilitate signal reception. The occurrence and quantity of pied tamarin' long calls, as well as their acoustic proprieties, seem to be governed by anthropogenic noise, time of the day, and social mechanisms such as proximity to neighboring groups.

Systematic review of machine learning methods applied to ecoacoustics and soundscape monitoring.

Soundscape ecology is a promising area that studies landscape patterns based on their acoustic composition. It focuses on the distribution of biotic and abiotic sounds at different frequencies of the landscape acoustic attribute and the relationship of said sounds with ecosystem health metrics and indicators (e.g., species richness, acoustic biodiversity, vectors of structural change, gradients of vegetation cover, landscape connectivity, and temporal and spatial characteristics). To conduct such studies, researchers analyze recordings from Acoustic Recording Units (ARUs). The increasing use of ARUs and their capacity to record hours of audio for months at a time have created a need for automatic processing methods to reduce time consumption, correlate variables implicit in the recordings, extract features, and characterize sound patterns related to landscape attributes. Consequently, traditional machine learning methods have been commonly used to process data on different characteristics of soundscapes, mainly the presence-absence of species. In addition, it has been employed for call segmentation, species identification, and sound source clustering. However, some authors highlight the importance of the new approaches that use unsupervised deep learning methods to improve the results and diversify the assessed attributes. In this paper, we present a systematic review of machine learning methods used in the field of ecoacoustics for data processing. It includes recent trends, such as semi-supervised and unsupervised deep learning methods. Moreover, it maintains the format found in the reviewed papers. First, we describe the ARUs employed in the papers analyzed, their configuration, and the study sites where the datasets were collected. Then, we provide an ecological justification that relates acoustic monitoring to landscape features. Subsequently, we explain the machine learning methods followed to assess various landscape attributes. The results show a trend towards label-free methods that can process the large volumes of data gathered in recent years. Finally, we discuss the need to adopt methods with a machine learning approach in other biological dimensions of landscapes.

Dataset for polyphonic sound event detection tasks in urban soundscapes: The synthetic polyphonic ambient sound source (SPASS) dataset.

This paper presents the Synthetic Polyphonic Ambient Sound Source (SPASS) dataset, a publicly available synthetic polyphonic audio dataset. SPASS was designed to train deep neural networks effectively for polyphonic sound event detection (PSED) in urban soundscapes. SPASS contains synthetic recordings from five virtual environments: park, square, street, market, and waterfront. The data collection process consisted of the curation of different monophonic sound sources following a hierarchical class taxonomy, the configuration of the virtual environments with the RAVEN software library, the generation of all stimuli, and the processing of this data to create synthetic recordings of polyphonic sound events with their associated metadata. The dataset contains 5000 audio clips per environment, i.e., 25,000 stimuli of 10 s each, virtually recorded at a sampling rate of 44.1 kHz. This effort is part of the project ``Integrated System for the Analysis of Environmental Sound Sources: FuSA System'' in the city of Valdivia, Chile, which aims to develop a system for detecting and classifying environmental sound sources through deep Artificial Neural Network (ANN) models.

Assessing the Soundscape Appropriateness in the Vicinity of a Heliport in an Urban Park of Quito (Ecuador) Using Immersive Audio-Visual Scenarios.

Heliports are facilities that play a fundamental role in security and emergency operations. Since rotorcrafts do not need much space for take-off and landing, heliports are normally immersed in the urban fabric of our cities. However, they generate high noise levels, which can cause a nuisance, especially in outdoor areas intended for the recreation of citizens. This paper studies how helicopter noise affects the perception of the soundscape appropriateness and landscape quality in the vicinity of a heliport located in an urban park, using semantic differential scales and appraisals on the noise sources. The study area was the "Parque del Bicentenario" in Quito, Ecuador. Immersive Virtual Reality (IVR) laboratory tests using 360-degree videos and spatial audio were preferred to on-site questionnaires, given the difficulty of predicting when helicopter noise events would occur. For the statistical analysis, objective acoustic and psychoacoustic parameters have also been considered. Results show that the soundscape is perceived as more pleasant and less chaotic when there is no helicopter noise. Furthermore, with the same visual stimuli, the appraisals of the landscape are much better in the scenarios without the helicopter noise. Sharpness is the psychoacoustic parameter that best explains the variance of the subjective variables evaluated.

The influence of anthropogenic noise on the behavior of male mantled howler monkeys.

Anthropogenic noise is increasingly disturbing natural soundscapes and affecting the physiology, behavior, and fitness of wildlife. However, our knowledge about the impact of anthropogenic noise on wild primates is scant. Here, we assess the effects of anthropogenic noise on the behavior of male mantled howler monkeys (Alouatta palliata). Specifically, we describe the types, rates, and sound pressure level (SPL) of anthropogenic noise that occurs in areas inhabited by mantled howler monkeys and determine if the behavioral responses of males to anthropogenic noise are influenced by noise attributes. For 1 year (1753 h), we characterized anthropogenic noise in the Los Tuxtlas Biosphere Reserve (Veracruz, Mexico) and studied the behavior of males belonging to five groups. Anthropogenic noise was common, diverse, and varied among areas in terms of rate, type, and SPL. Males did not display behavioral responses toward most (60%) anthropogenic noises, but were more likely to respond to certain noise types (e.g., aerial traffic) and toward noise with high SPL. Group identity influenced the likelihood of displaying behavioral responses to noise. The most common behavioral responses were vocalizations and vigilance. Males vocalized in response to noise with high SPL, although this relationship depended on group identity. The effect of the number of noises on vocalizations also varied among groups. Males were more likely to display vigilance toward high SPL and infrequent noise, but, again, these relationships varied among groups. In sum, anthropogenic noise is pervasive in areas inhabited by mantled howler monkeys and influences male behavior. Experience and frequency of exposure may modulate the behavioral responses of male mantled howler monkeys to noise and explain the group differences.

Development and application of practical criteria for the recognition of potential Health Restoration Soundscapes (HeReS) in urban greenspaces.

Urban greenspace soundscapes can contribute to the restorative effects that nature provides for the psychological and emotional health of people when certain conditions are met. The main objective of this paper is to propose practical criteria to help planners and managers in the design, development and preservation of urban greenspaces whose soundscapes may contribute to the renewal of health. Systematic literature review informed a conceptual potential Health Restoration Soundscapes (HeReS) model, based on five conditions: (1) Naturalness, (2) Sound Levels, (3) Perceived Sound Sources, (4) Soundscape Assessment, and (5) Sensescape Coherence and proposed Health Restoration Soundscapes Criteria (HeReS-C), for HeReS evaluation in urban green areas. The HeReS-C were applied in 21 sites in Argentina, Sweden, and Chile. General results are provided for all 21 sites, along with three in-depth profiles of HeReS-C applications that provide case studies across a range of resulting HeReS-C scenarios, including sites that meet the HeReS-C criteria, those that do not, and sites that could qualify in the future, if appropriate management measures are taken. HeReS-C showed to be a promising tool for the recognition of potential health-restoring soundscapes in urban greenspaces; informing their design and management to favor the well-being and health of the population.

Traffic Noise Annoyance in the Population of North Mexico: Case Study on the Daytime Period in the City of Matamoros.

Aim: The presence of noise in urban environments is rarely considered a factor that causes damage to the environment. The primary generating source is transportation means, with vehicles being the ones that affect cities the most. Traffic noise has a particular influence on the quality of life of those who are exposed to it and can cause health alterations ranging from annoyance to cardiovascular diseases. This study aims to describe the relationship between the traffic noise level and the perceived annoyance in the inhabitants of a city on the Northern Border of Mexico. The work carried out in a city represents the vulnerability characteristics: economic, social, and migratory of its sizable portion of the inhabitants. Due to that, it is impossible to identify precisely the number of residents as the number of vehicles in circulation. Methods: The streets and avenues with an annual average daily traffic of more than 1,000 vehicles were considered for the measurement of traffic noise. The equipment used was a vehicle gauge with non-invasive speed radar; type I integrating sound level meters, with their respective gauges and tripods. A questionnaire was applied to people living within 250 m of the streets and avenues in which the noise was measured. Results: The noise measurement found a parameter of LAeq estimated for 12 h during the day, exceeding 70 dBA. The data received from the questionnaire were statistically tested by using Pearson's correlation tests. A total number of 2,350 people were participated, of whom 1,378 were women (58.6%) and 972 were men (41.4%). The age of participants is ranged from 18 to 75 years. The overall perception of traffic noise annoyance identified that 1,131 participants (48.1%) responded "Yes" as they considered the noise annoying. Participants who responded "No" as well as those who responded "Do not know" resulted in a total of 1,219 people (51.9%). Conclusion: The results show that the population is desensitized to traffic noise and does not perceive it as an annoyance. The flow of vehicles and the type of vehicles are the significant factors for the propagation and increase in the traffic noise levels. Women present a considerable appreciation of traffic noise perception instead of younger people who demonstrate a higher tolerance to high-level exposure. This reflects the lack of information of the population around the noise problem and its effects.

Influence of natural and anthropogenic sound sources on the soundscape of the Cananéia estuary, southeastern Brazil.

The aim of this study is to characterize the background noise and abiotic and anthropogenic sound sources in the sector with greater anthropogenic use of the Cananéia estuary. The results show that the relative amplitude of background noise decreased with the increase of frequency range, and was higher with greater number of vessels, wind speed and during flood tide. Weekends and vacation periods were shown to be important dates during which background noise increased in the region. The influence of the tide and the wind speed on the relative amplitude was dependent on the frequency range analyzed. Therefore, both abiotic and anthropogenic sound sources were observed to be important factors regarding an increase in background noise in Cananéia. The importance of the continuity of vessel regulation in the region and of future studies that identify whether such noises alter parameters from the sound repertoire of the cetacean species in the region.

On the Person-Place Interaction and Its Relationship with the Responses/Outcomes of Listeners of Urban Soundscape (Compared Cases of Lisbon and Bogotá): Contextual and Semiotic Aspects.

Design, planning, and management of the urban soundscape require various interacting fields of knowledge given the fact that it is the human person that experiences and provides meaning to the urban places and their acoustic environments. The process of environmental perception involves contextual information that conditions people's responses and outcomes through the relationship between the variables Person, Activity, and Place. This research focuses on the interaction between Person and Place and its impact on responses and outcomes from listeners with different geographical origin and background. Laboratory studies were conducted in the cities of Lisbon (Portugal) and Bogotá (Colombia), where local listeners were introduced to known and unknown acoustic environments. Sound data recorded in the two cities allowed comparison of responses and outcomes of the listeners according to the Person-Place Interaction, leading to different meanings depending on the contextual variables. The results clearly show a relationship between site, acoustic environment, soundscape, Person-Place Interaction, and meaning of the place. This information can be useful for urban technicians and designers dealing with planning and management of urban soundscapes.

A Green Soundscape Index (GSI): The potential of assessing the perceived balance between natural sound and traffic noise.

Urban soundscapes are dynamic and complex multivariable environmental systems. Soundscapes can be organized into three main entities containing the multiple variables: Experienced Environment (EE), Acoustic Environment (AE), and Extra-Acoustic Environment (XE). This work applies a multidimensional and synchronic data-collecting methodology at eight urban environments in the city of Córdoba, Argentina. The EE was assessed by means of surveys, the AE by acoustic measurements and audio recordings, and the XE by photos, video, and complementary sources. In total, 39 measurement locations were considered, where data corresponding to 61 AE and 203 EE were collected. Multivariate analysis and GIS techniques were used for data processing. The types of sound sources perceived, and their extents make up part of the collected variables that belong to the EE, i.e. traffic, people, natural sounds, and others. Sources explaining most of the variance were traffic noise and natural sounds. Thus, a Green Soundscape Index (GSI) is defined here as the ratio of the perceived extents of natural sounds to traffic noise. Collected data were divided into three ranges according to GSI value: 1) perceptual predominance of traffic noise, 2) balanced perception, and 3) perceptual predominance of natural sounds. For each group, three additional variables from the EE and three from the AE were applied, which reported significant differences, especially between ranges 1 and 2 with 3. These results confirm the key role of perceiving natural sounds in a town environment and also support the proposal of a GSI as a valuable indicator to classify urban soundscapes. In addition, the collected GSI-related data significantly helps to assess the overall soundscape. It is noted that this proposed simple perceptual index not only allows one to assess and classify urban soundscapes but also contributes greatly toward a technique for separating environmental sound sources.

A comprehensive methodology for the multidimensional and synchronic data collecting in soundscape.

The soundscape paradigm is comprised of complex living systems where individuals interact moment-by-moment among one another and with the physical environment. The real environments provide promising conditions to reveal deep soundscape behavior, including the multiple components involved and their interrelations as a whole. However, measuring and analyzing the numerous simultaneous variables of soundscape represents a challenge that is not completely understood. This work proposes and applies a comprehensive methodology for multidimensional and synchronic data collection in soundscape. The soundscape variables were organized into three main entities: experienced environment, acoustic environment, and extra-acoustic environment, containing, in turn, subgroups of variables called components. The variables contained in these components were acquired through synchronic field techniques that include surveys, acoustic measurements, audio recordings, photography, and video. The proposed methodology was tested, optimized, and applied in diverse open environments, including squares, parks, fountains, university campuses, streets, and pedestrian areas. The systematization of this comprehensive methodology provides a framework for soundscape research, a support for urban and environment management, and a preliminary procedure for standardization in soundscape data collecting.

Underwater soundscape of marine protected areas in the south Brazilian coast.

The Laje de Santos Marine State Park (LSMSP) and Xixová-Japuí State Park (XJSP) are two protected areas (PA), close to the Santos Bay in the south Brazilian coast. The region encompasses both important biodiversity and anthropogenic activities. This study aims to serve as a first reference survey of the underwater soundscape of these PAs. Additionally it evaluates the presence of the anthropogenic and biological sound in these areas. One month of continuous recorded underwater sound, at selected locations in XJSP and LSMSP, is used in this study. The data were characterized by its spectral content and by the temporal evolution of Sound Pressure Levels (SPL). Both locations showed sound events with daily periodicities, mainly related with boats and fish chorus.

Ambient noise and temporal patterns of boat activity in the US Virgin Islands National Park.

Human activity is contributing increasing noise to marine ecosystems. Recent studies have examined the effects of boat noise on marine fishes, but there is limited understanding of the prevalence of this type of sound source. This investigation tracks vessel noise on three reefs in the US Virgin Islands National Park over four months in 2013. Ambient noise levels ranged from 106 to 129dBrms re 1µPa (100Hz-20kHz). Boat noise occurred in 6-12% of samples. In the presence of boat noise, ambient noise in a low-frequency band (100-1000Hz) increased by >7dB above baseline levels and sound levels were significantly higher. The frequency with the most acoustic energy shifted to a significantly lower frequency when boat noise was present during the day. These results indicate the abundance of boat noise and its overlap with reef organism sound production, raising concern for the communication abilities of these animals.