Automatic Acoustic Monitoring

We report on the development of an automated acoustic bird recognizer with imp roved noise robustness, which is part of a long-term project, aiming at the establishment of an automated biodiversity monitoring system at the Hy mettus Mountain near Athens, Greece. In particu lar, a typical audio p rocessing strategy, which has been proved quite successful in various audio recognition applications, was amended with a simp le and effective mechanis m fo r integration of temporal contextual information in the decisionmaking process. In the present implementation, we consider integration of temporal contextual information by joint post-processing of the recognition results for a number of preceding and subsequent audio frames. In order to evaluate the usefulness of the proposed scheme on the task of acoustic bird recognition, we experimented with six widely used classifiers and a set of real-field audio record ings for two bird species which are present at the Hy mettus Mountain. The highest achieved recognition accuracy obtained on the real-field data was approximately 93%, while experiments with additive noise showed significant robustness in low signal-to-noise ratio setups. In all cases, the integration of temporal contextual informat ion was found to improve the overall accuracy of the recognizer.

Road and rail tunnels face risks from structural degradation (cracks, corrosion, etc.) and environmental stressors (flooding, seismic activity, etc.), yet conventional SHM systems often react post damage, lacking preventive capabilities. Conventional SHM systems, reliant on grid power, require extensive maintenance, increasing costs and limiting reliability in remote sections. This research proposes a "smart ear" system for preventive monitoring, using audio sensors (piezoelectric transducers) and predictive AI to enhance tunnel resilience. AI models, including RNNs and transformers, forecast risks before escalation, enabling proactive maintenance. Piezoelectric transducers harvest energy from traffic vibrations, ensuring sustainability. A digital twin integrates audio data to simulate tunnel behavior, optimizing preventive strategies. Targeting seismic and flood-prone regions globally, the system will be validated in lab and real-world international cases. Objectives include system design, AI development, and evaluation of safety, cost, and sustainability impacts. The approach is novel for its acoustic-based monitoring, predictive AI, and self-powered design. Hypothesis: The system will achieve 95% precision, 85% recall, and 88% mAP in risk prediction, with a 10 cm² piezoelectric transducer generating 100–500 µW at 1–5 V and 100–500 µA, able to power the monitoring device continuously for 30 days without external charging.

We report on the development of an automated acoustic bird recognizer with imp roved noise robustness, which is part of a long-term project, aiming at the establishment of an automated biodiversity monitoring system at the Hy mettus Mountain near Athens, Greece. In particu lar, a typical audio p rocessing strategy, which has been proved quite successful in various audio recognition applications, was amended with a simp le and effective mechanis m fo r integration of temporal contextual information in the decisionmaking process. In the present implementation, we consider integration of temporal contextual information by joint post-processing of the recognition results for a number of preceding and subsequent audio frames. In order to evaluate the usefulness of the proposed scheme on the task of acoustic bird recognition, we experimented with six widely used classifiers and a set of real-field audio record ings for two bird species which are present at the Hy mettus Mountain. The highest achieved recognition accuracy obtained on the real-field data was approximately 93%, while experiments with additive noise showed significant robustness in low signal-to-noise ratio setups. In all cases, the integration of temporal contextual informat ion was found to improve the overall accuracy of the recognizer.

Although acoustic recordings have recently gained popularity as an alternative to point counts for surveying birds, little is known about the relative performance of the two methods for detecting tropical bird species across multiple vegetation types. During June and July 2008, we collected species detection/nondetection data to compare the performance of a quadraphonic acoustic recording system and point counts for estimating species richness and composition and detection probabilities of 15 rare, moderately common, and common tropical bird species across six structurally distinct vegetation types (coastal dune scrub, mangrove, low-stature deciduous thorn forest, early and late successional medium-stature semievergreen forest, and grazed pastures) in the northern Yucatan Peninsula. We selected five rare species endemic to the Yucatan Peninsula and 10 moderately common and common species that also occur in other tropical regions. Species richness and composition did not differ between survey methods in any of the vegetation types. At the population level, however, we found support for an effect of method on detection probability for most species. For 13 species, regardless of their abundance, acoustic recordings yielded detection probabilities as high as or higher than those for point counts across all vegetation types. The remaining two species were better detected by point counts in pastures and coastal scrub, where greater visibility likely improved sightings of these species. However, these species were detected as well as or better by acoustic recordings in forests and mangroves where detections were primarily auditory. In tropical regions where experienced field observers may not be available and funding for field surveys may be limited, acoustic recordings offer a practical solution for determining species richness and composition and the occupancy patterns of most species. However, for some species, a combination of methods will provide the most reliable data. Regardless of the method selected, analyses that account for variation in detection probability among vegetation types will be necessary because most species in our study demonstrated vegetation-dependent detection probabilities. RESUMEN. Eficacia y utilidad de las grabaciones acústicas para el muestreo de aves tropicales El uso de grabaciones acústicas recientemente ha ganado popularidad como una alternativa a los conteos por punto para el muestreo de aves. Sin embargo, poco se sabe sobre el desempeño de ambos métodos en la detección

AMIBIO AmiBio concept: The AmiBio project constructs and deploys autonomous multi-sensor monitoring stations and automatically analyzes their transmitted meas-urements, for long-term monitoring of biodiversity activity trends in the region of Hymettus, a mountainous area located at the eastern periphery of Athens. The acoustic monitoring stations, which are non-intrusive and cost-effective compared to human expeditions, will assist in systematic seasonal and longitudi-nal long-term environmental monitoring that will allow the automatic inventory and examination of the biological diversity of the study area. The monitoring stations will record audio signals, temperature, humidity, illumi-nation, wind speed and direction and transmit these data to a central station far away from the monitoring region. Special computer software at the central sta-tion, deployed in the premises of the Association for Protection and Develop-ment of Hymettus (SPAY) and in the University of Patras, will th...

review my work and provided valuable comments on this research. Special thanks to my wife, Jill Deppe, who helped me on every step of this project. Thanks to the organizations who provided me with financial support for my research: Association of Field Ornithologists (Alexander Bergstrom Award), Cleveland Metroparks Zoo and Cleveland Zoological Society (Scott Neotropical Fund), Idea Wild, and the Illinois Natural History Survey. This project would not have been possible without the logistical support of the El Eden Ecological Reserve, the Ria Lagartos Biosphere Reserve and many volunteers and friends who assisted me in the field. I also thank T. J. Benson for his suggestions and willingness to always answer my questions. Finally, thanks to all my family and numerous friends who supported me during this exciting process.

The AmiBio project constructs and deploys autonomous multisensor monitoring stations and automatically analyzes their transmitted measurements, for long-term monitoring of biodiversity activity trends in the region of Hymettus, a mountainous area located at the eastern periphery of Athens. The acoustic monitoring stations, which are non-intrusive and cost-effective compared to human expeditions, will assist in systematic seasonal and longitudinal long-term environmental monitoring that will allow the automatic inventory and examination of the biological diversity of the study area. The monitoring stations will record audio signals, temperature, humidity, illumination, wind speed and direction and transmit these data to a central station far away from the monitoring region. Special computer software at the central station, deployed in the premises of the Association for Protection and Development of Hymettus (SPAY) and in the University of Patras, will then perform automatic species identification and statistical analysis, thus contributing to a continuous biodiversity survey of the designated areas. The project team plans to deploy the prototype hardware system in relevant habitats, such as forest, open habitats, and rock cliffs, that will gather and transmit recordings of audio and data on environmental variables to a central station (see Fig. ).  Data repository of audio recordings and database management tools in support of research on bioacoustics and animal behaviour.  Software for automatic recognition of species from their sound emissions.  Prototype of an acoustic monitoring system operating in 24/7 mode.  Estimation of the density of sound-emitting animal populations.  Monitoring of rare and threatened species.  Assessment of the conservation status of the monitored sites.  Monitoring and alarming of specific atypical sound events, such as gunshots, off-road motor bikers, etc.  Monitoring for natural calamity and human induced disasters. At present, the prototype remote monitoring stations and the supporting database tools and software for bioacoustic recognition of species are implemented, and the field tests have started -further details are provided on pages 4-7.

The on-going AmiBio project † (www.amibio-project.eu) aims to construct and deploy autonomous multi-sensor surveillance stations and to automatically analyze the transmitted data for the long-term monitoring of biodiversity activity trends in the Natura 2000 area "Hymettus Mountains", located east of Athens,

Although acoustic recordings have recently gained popularity as an alternative to point counts
for surveying birds, little is known about the relative performance of the two methods for detecting tropical bird
species across multiple vegetation types. During June and July 2008, we collected species detection/nondetection
data to compare the performance of a quadraphonic acoustic recording system and point counts for estimating
species richness and composition and detection probabilities of 15 rare, moderately common, and common tropical
bird species across six structurally distinct vegetation types (coastal dune scrub, mangrove, low-stature deciduous
thorn forest, early and late successional medium-stature semievergreen forest, and grazed pastures) in the northern
Yucatan Peninsula. We selected five rare species endemic to the Yucatan Peninsula and 10 moderately common and
common species that also occur in other tropical regions. Species richness and composition did not differ between
survey methods in any of the vegetation types. At the population level, however, we found support for an effect of
method on detection probability for most species. For 13 species, regardless of their abundance, acoustic recordings
yielded detection probabilities as high as or higher than those for point counts across all vegetation types. The
remaining two species were better detected by point counts in pastures and coastal scrub, where greater visibility
likely improved sightings of these species. However, these species were detected as well as or better by acoustic
recordings in forests and mangroves where detections were primarily auditory. In tropical regions where experienced
field observers may not be available and funding for field surveys may be limited, acoustic recordings offer a practical
solution for determining species richness and composition and the occupancy patterns of most species. However,
for some species, a combination of methods will provide the most reliable data. Regardless of the method selected,
analyses that account for variation in detection probability among vegetation types will be necessary because most
species in our study demonstrated vegetation-dependent detection probabilities.

Concept † The AmiBio project (LIFE08 NAT/GR/000539) is implemented with the contribution of the LIFE financial instrument of the European Community. Duration: 01−FEB−2010 to 30−JUN −2013. Total Budget: 1,664,282 €. Life+ Contribution: 830,641 €.

AMIBIO AmiBio concept: The AmiBio project constructs and deploys autonomous multi-sensor monitoring stations and automatically analyzes their transmitted meas-urements, for long-term monitoring of biodiversity activity trends in the region of Hymettus, a mountainous area located at the eastern periphery of Athens. The acoustic monitoring stations, which are non-intrusive and cost-effective compared to human expeditions, will assist in systematic seasonal and longitudi-nal long-term environmental monitoring that will allow the automatic inventory and examination of the biological diversity of the study area. The monitoring stations will record audio signals, temperature, humidity, illumi-nation, wind speed and direction and transmit these data to a central station far away from the monitoring region. Special computer software at the central sta-tion, deployed in the premises of the Association for Protection and Develop-ment of Hymettus (SPAY) and in the University of Patras, will then perform auto-matic species identification and statistical analysis, thus contributing to a con-tinuous biodiversity survey of the designated areas. The project team plans to deploy the prototype hardware system in relevant habitats, such as forest, open habitats, and rock cliffs, that will gather and transmit recordings of audio and data on environmental variables to a central station (see Fig. 1). Expected outcomes from the project:  Data repository of audio recordings and database management tools in sup-port of research on bioacoustics and animal behaviour. Software for automatic recognition of species from their sound emissions. Prototype of an acoustic monitoring system operating in 24/7 mode.  Estimation of the density of sound-emitting animal populations.  Monitoring of rare and threatened species.  Assessment of the conservation status of the monitored sites. Monitoring and alarming of specific atypical sound events, such as gunshots, off-road motor bikers, etc. Monitoring for natural calamity and human induced disasters.

We report on the development of an automated acoustic bird recognizer with improved noise robustness, which is part of a long-term project, aiming at the establishment of an automated biodiversity monitoring system at the Hymettus Mountain near Athens, Greece. In particular, a typical audio processing strategy, which has been proved quite successful in various audio recognition applications, was amended with a simple and effective mechanism for integration of temporal contextual information in the decisionmaking process. In the present implementation, we consider integration of temporal contextual information by joint post-processing of the recognition results for a number of preceding and subsequent audio frames. In order to evaluate the usefulness of the proposed scheme on the task of acoustic bird recognition, we experimented with six widely used classifiers and a set of real-field audio recordings for two bird species which are present at the Hymettus Mountain. The highest achieved recognition accuracy obtained on the real-field data was approximately 93%, while experiments with additive noise showed significant robustness in low signal-to-noise ratio setups. In all cases, the integration of temporal contextual information was found to improve the overall accuracy of the recognizer.

In other words, for these sites it is reasonable to assume that the majority of tortoises detected by the soundrecording stations pertain to the former species, not to T. hermanni. The Ergani Valley obviously is an ideal reproductive site for the Marginated Tortoise, which satisfies the ecological requirements of the species, such as high population density of reproductive adults, abundant food resources, nest sites, and shelter. Besides being an important breeding site for this tortoise, the valley provides habitats for several other animal species of conservation concern, particularly birds (e.g., Lesser Kestrel Falco naumanni, Peregrine Falcon Falco peregrinus, Short-toed Snake-Eagle Circaetus gallicus, Long-legged Buzzard Buteo rufinus, Red-backed Shrike Lanius collurio, and Cretzschmar's Bunting Emberiza caesia) and the European Green Toad Pseudepidalea viridis. Ergani currently is affected by strong anthropogenic pressures due to uncontrolled dumping of construction waste and illegal motocross activities, causing an immense negative environmental impact. This situation suggests the urgent need of a conservation action plan for the area to maintain favorable ecological conditions for species protected under EU legislation. Cover photo: Couple of Marginated Tortoise, photographed by ZFMK biology students, M. Konz and L. Lemm, in the Hymettus area.