Mobile Technology Helps Monitor Biodiversity
| Autor: | Diep, Daniel, Nonon, Hervé, Marc, Isabelle, Delhom, Jonathan, Roure, Frederic |
|---|---|
| Přispěvatelé: | IMT - Mines Alès, Administrateur, Laboratoire de Génie Informatique et Ingénierie de Production (LGI2P), IMT - MINES ALES (IMT - MINES ALES), Institut Mines-Télécom [Paris] (IMT)-Institut Mines-Télécom [Paris] (IMT) |
| Jazyk: | angličtina |
| Rok vydání: | 2013 |
| Předmět: | |
| Zdroj: | IBAC 2013-International Bioacoustics Congress IBAC 2013-International Bioacoustics Congress, Sep 2013, Pirenópolis, Brazil |
| Popis: | International audience; Smartphones and tablet computers provide today a substantial computing power combined with high communication facilities, and they can constitute a low cost, small size base platform to monitor the behaviour of animal species. In the work presented here, a smartphone has been used to estimate populations of shad fish in rivers, by analyzing acoustic signals. Twaite shad (Alosa fallax) is a migratory fish living primarily in the seas which goes up the rivers to breed in spring. In Europe, this species has considerably declined due to overfishing, pollution and obstacles to migration, and is now given considerable legal protection. Monitoring the numbers of shads at their reproduction sites is an important indicator for measuring interannual changes in their population. Shad reproduce at night, emitting a characteristic sound lasting a few seconds known as a "spawning splash". In this paper, we present the design of a new equipment achieved to automatically count and record the spawning splashes at a reproduction site. The central part of this equipment is a smartphone, and this configuration enables to benefit from all the functions attached to it: audio recording capability, high storage capacity, wireless communication, power autonomy, the whole being integrated in a small size device. Audio acquisition by an external microphone was completed with a parabolic reflector. Automatic detection has been decomposed in two steps: in the first step, characteristic features are extracted from the acoustic signals in order to provide a representation of the signals in a space of reduced dimension. 10 spectral parameters have been used for each time period of 93 ms. The second step consists in classifying the signals and detecting spawning splashes, on the basis of a training phase with recorded data. A Gaussian Mixture Model was trained to determine clusters of splash signals. First test results obtained in 2012 show a correct detection rate of more than 80%. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|