Dataset Information

An Efficient Model for a Vast Number of Bird Species Identification Based on Acoustic Features.

ABSTRACT: Birds have been widely considered crucial indicators of biodiversity. It is essential to identify bird species precisely for biodiversity surveys. With the rapid development of artificial intelligence, bird species identification has been facilitated by deep learning using audio samples. Prior studies mainly focused on identifying several bird species using deep learning or machine learning based on acoustic features. In this paper, we proposed a novel deep learning method to better identify a large number of bird species based on their call. The proposed method was made of LSTM (Long Short-Term Memory) with coordinate attention. More than 70,000 bird-call audio clips, including 264 bird species, were collected from Xeno-Canto. An evaluation experiment showed that our proposed network achieved 77.43% mean average precision (mAP), which indicates that our proposed network is valuable for automatically identifying a massive number of bird species based on acoustic features and avian biodiversity monitoring.

SUBMITTER: Wang H

PROVIDER: S-EPMC9495150 | biostudies-literature | 2022 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

An Efficient Model for a Vast Number of Bird Species Identification Based on Acoustic Features.

Wang Hanlin H Xu Yingfan Y Yu Yan Y Lin Yucheng Y Ran Jianghong J

Animals : an open access journal from MDPI 20220915 18

Birds have been widely considered crucial indicators of biodiversity. It is essential to identify bird species precisely for biodiversity surveys. With the rapid development of artificial intelligence, bird species identification has been facilitated by deep learning using audio samples. Prior studies mainly focused on identifying several bird species using deep learning or machine learning based on acoustic features. In this paper, we proposed a novel deep learning method to better identify a l ...[more]

PMID: 36139299

Similar Datasets

Project description:BackgroundTargeted PCR amplicon sequencing (TAS) techniques provide a sensitive, scalable, and cost-effective way to query and identify closely related bacterial species and strains. Typically, this is accomplished by targeting housekeeping genes that provide resolution down to the family, genera, and sometimes species level. Unfortunately, this level of resolution is not sufficient in many applications where strain-level identification of bacteria is required (biodefense, forensics, clinical diagnostics, and outbreak investigations). Adding more genomic targets will increase the resolution, but the challenge is identifying the appropriate targets. VaST was developed to address this challenge by finding the minimum number of targets that, in combination, achieve maximum strain-level resolution for any strain complex. The final combination of target regions identified by the algorithm produce a unique haplotype for each strain which can be used as a fingerprint for identifying unknown samples in a TAS assay. VaST ensures that the targets have conserved primer regions so that the targets can be amplified in all of the known strains and it also favors the inclusion of targets with basal variants which makes the set more robust when identifying previously unseen strains.ResultsWe analyzed VaST's performance using a number of different pathogenic species that are relevant to human disease outbreaks and biodefense. The number of targets required to achieve full resolution ranged from 20 to 88% fewer sites than what would be required in the worst case and most of the resolution is achieved within the first 20 targets. We computationally and experimentally validated one of the VaST panels and found that the targets led to accurate phylogenetic placement of strains, even when the strains were not a part of the original panel design.ConclusionsVaST is an open source software that, when provided a set of variant sites, can find the minimum number of sites that will provide maximum resolution of a strain complex, and it has many different run-time options that can accommodate a wide range of applications. VaST can be an effective tool in the design of strain identification panels that, when combined with TAS technologies, offer an efficient and inexpensive strain typing protocol.

Project description:BackgroundThe potentially negative effects of timber harvesting on biodiversity and habitat conservation leads to the consideration of a wide range of restrictions to forest logging in natural areas. In particular, high noise levels produced by forest machinery present a challenge to developing sustainable forest management plans. The Cinereous vulture (Aegypius monachus), the largest bird of prey whose nests are located in mature trees, is considered to be appropriate as an indicator species for environment-friendly forest planning. In this work, we evaluated spatially differences in sound propagation between stands. We hypothesized that differences due to the influence of orography in mountainous forests would allow the relaxation of spatial and temporary restrictions to timber logging, without causing any great disturbance at nesting sites of sensitive species.MethodsOur study was conducted in a Scots pine (Pinus sylvestris) forest of Spain, where an important colony of the Western European population of Cinereous vulture is located. We built 62 noise maps to characterize noise pollution due to tree logging at planning level. We modeled two different scenarios, in order to characterize; (i) the effect of a chainsaw operator during a complete cycle for felling a tree (Scenario 1), and (ii) the effect of the peak level produced by the breaking noise emitted by the trunk of the tree and its impact on the ground (Scenario 2). A strategy of three logical steps was designed; (i) landscape-scale analysis of noise propagation in stands, (ii) hierarchical cluster analysis of stands, (iii) assessment of the potentially significant influence of noise management in timber harvesting.ResultsThe minimum distance (DIST) from chainsaw operator sites to the 40 dB(A) contour lines was the only variable that had a significant influence on the clustering results. On the other hand, mean values of a newly proposed metric called average radius or radial distance (ARD) oscillated between 174 m in cluster #1 (Scenario 1) and 407 m in cluster #2 (Scenario 2).DiscussionOur results highlight the convenience of considering noise modeling tools at a forest planning level in order to address the compatibility of forest logging and the necessity of protecting nature. We found that spatial propagation of noise made by chainsaws at felling sites does not differ between stands even in a mountainous terrain, contrary to what we initially hypothesized. However, temporary logging restrictions could be excluded in about 36% of the current conditioned management areas according to ARD results in Scenario 2 (400 m). This proposal could be based on a sound pressure level (in decibels) criteria instead of conventional buffer protection distance criteria. In addition, it is suggested that the current size of restricted management areas could be generally extended from a 100 m radius to a 200 m one around the Cinereous vulture nest sites.

Dataset Information

An Efficient Model for a Vast Number of Bird Species Identification Based on Acoustic Features.

Publications

An Efficient Model for a Vast Number of Bird Species Identification Based on Acoustic Features.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets