Themes of this team
Communication Systems in constraining environments
Codage des signaux et spéciation
The research activity of The Bioacoustics Team is centred on animal acoustic communication.
Several thematics are developed including :
1- Communication in constraining environments
2- Sexual Selection
3- Neuro-Ethology of acoustic communication
4- Communication Networks
Communication Systems in constraining environments
In animals, recognition between individuals is essential to the settlement of sexual and social relationships. Due to their physical properties and their potentiality to encode any kind of information, sounds are an effective mean to reliably transmit the identity of the emitter. Nevertheless, in colonial animals, the vocal signal an adult produces when seeking its young or its partner among thousands of individuals is transmitted in a context involving the noise generated by the colony. This background noise drastically reduces the signal-to-noise ratio and masks the signal by sounds with similar spectral and temporal characteristics. We investigate in colonial bird and mammal species how this extreme acoustic environment constrains the transfer of acoustic information. We examine solutions found at the level of the emitter to improve the efficiency of communication and we report how the receiver can optimize the collected information.
Studied Models : penguins, gulls, pinnipeds, sheep.
Dense forests represent constraining acoustic channels. In this obstructed environment, sounds are modified during propagation by various processes such attenuation, reverberations, frequency filtering and scattering. In spite of these propagation-induced alterations of the sound, acoustic communication remains effective. The aim of our research is first to characterize the modifications of acoustic signals during transmission, and second to understand how birds manage with these environmental constraints to communicate at long rate. For a comparison purpose, we focus both on temperate and tropical forest songbirds.
Studied models : white-brown warbler, blackcap wren, birds of the rain forest.
SPECIES, GROUP ANS INDIVIDUAL ACOUSTIC SIGNATURES
Vocal signals are involved in many social interactions and can encode several types of information such as species identity, group label, gender, social status and individual identity. Such information is essential for reproduction success and survival, but also to organize social links. We investigate the coding-decoding processes of this information. Coding of information is studied by performing exhaustive acoustic analyses, but the most interesting part is to decrypt which particular acoustic parameters (FM, AM, timbre…) are used by animals for these different types of recognition. To decode these “species”, “group” or “individual” vocal signatures, we synthesized calls in which one given parameter has been modified, and we “ask” animals with playback experiments if they still recognize the modified calls as coming from its own species or its own group or from its mate or young.
Studied models : amphibians, crocodilians, penguins, pinnipeds, seabirds, songbirds.
SOCIAL STRUCTURES AND INDIVIDUAL RECOGNITION
We investigate how communication strategies involved with social structures and breeding ecology. Among species phylogenically related, they can show different social structures, from solitary to highly colonial species, inducing different constraints interfering with their communication system. By studying different species among birds and mammals, we are willing to demonstrate that communication systems are mainly linked to ecological rather than genetic traits.
Studied models : penguins, pinnipeds, seabirds.
Animals show a vast diversity of vocalizations from one species to another and from one individual to another within a species. This can be problematic in analysing vocalizations. The Fourier spectrogram remains today the classical time-frequency tool used by biologists – and sometimes the only one proposed – for use with typical software dedicated to bioacoustic sound analysis, such as MobySoft Ishmael, RainbowClick, Raven, Avisoft, and XBat.
The spectrogram is sufficient for many cases but we are working on recent signal processing methods : parametric linear models such as AutoRegressive filters, Schur algorithm, Wavelet transform, hilbert-huang transform. The objective is to give the best representation of these signals for extracting the pertinent information. The next step is to extract patterns for classification. The used methods are statistical approach, artificial neural netwoks, support vector machine, hidden markov chains.
Our projects are actually focussed on cetacean sounds. Especially, we work on sperm whales clicks, blue whales calls and humpback whales songs.
From numerous years, the team is implicated in bird scaring by mean of acoustic methods. Thus, we have patented an acoustic system to move some bird species (gulls, crows, lapwings, starlings, doves) away from runways. This system, based on the use of synthetic distress calls, is used with success by most of the important European airports. Currently, we are working on a method to reduce bird strike risk thanks to an acoustic system embedded on aircraft (Bird Impact Repellent & Deterrent sYstem R&T project with AIRBUS Industry group).
ACOUSTIC SURVEY AND CONSERVATION
Being not invasive and allowing investigation in dark or obstructed environments, acoustics can be a powerful tool of wildlife assessment. Currently, our activity focuses on two main questions :
1) surveying biodiversity in tropical rainforest and
2) following individuals within bird populations.
Concerning the first point, we work in Brazil in the Atlantic and Amazonian forests.
By focusing on species of interest, we are setting up systems of automatic recordings and recognition in order to assess the position of individuals of different studied species.
In regards to the second point, we are currently working on the woodcock Scolopax rusticola in the Azores Archipelago and the Great owl Bubo bubo in France.
Recently, we have started a study of an AAD (Automatic Acoustic Detection) system designed for pheasant population count in the field. The calls will be recorded by an automatic acoustic station and the collected data will be analyzed by a neural network in order estimate the number of individuals located in a given area. This study is realized in partnership with the Office National de la Chasse et de la Faune Sauvage (ONCFS).