National audienceThis paper presents new tools to improve a geoacoustic inversion scheme relying on the inversion of marine mammals vocalizations recorded on a single hydrophone. This method makes best use of the multipath propagation and time-frequency signature of vocal calls. The classical signal processing tools based on spectrogram limit both the range of the scheme and the signals panel that can be processed. In this paper, we introduce new tools based on temporal warping that allow high resolution paths separation. Each echo has to be transformed into a pure frequency. Therefore, they are warped according to the frequency law of the signal estimated according to the first echo. The warping tools enable a high resolution of each echo in the time-frequency warped domain. Applying the warping on a moving window, one can estimate the impulsive response (IR) of the channel and deduce the source location, grazing angle and transmission losses. Then, as warping operators conserve energy, the path levels can be estimated using the time-frequency warped representation. Through the processing of several signals, we get a curve of an estimation of the reflection coefficient that feeds the inversion algorithm. The theory of this extended method is described and its performances are evaluated on a controlled real data set in the Gulf of Lion. The range and the recording duration have been improved (range from 300 to 900 meters and duration multiplied by 2.5). With the new scheme, we have access to a set of geoacoustic parameters allowing to better describe the bottom features.Cet article introduit de nouveaux outils pour le traitement de signal dans le cadre d’un schéma d’inversion géoacoustique passive s’appuyant sur des modulations de fréquence de mammifères marins captés sur un unique hydrophone. La méthode se base sur l’utilisation de la propagation multitrajet pour extraire l’information nécessaire à l’estimation du coefficient de réflexion du sédiment superficiel. Jusqu’à présent, l’utilisation d’outils de traitement du signal basés sur le spectrogramme limitait la portée et les signaux candidats à l’inversion, ceux-ci devant présenter des échos résolus dans le plan temps-fréquence au regard des outils mis en œuvre. Ce papier propose l’ajout d’une nouvelle étape de traitement basée sur des opérateurs de déformation temporelle (warping) permettant la séparation haute résolution des échos. Les échos sont transformés en composantes de fréquences pures par application d’un opérateur de déformation avec pour loi de déformation, la loi de fréquence instantanée du signal. En contexte passif, cette loi est estimée à partir du premier écho reçu. L’application de l’opérateur de déformation sur une fenêtre temporelle glissante permet d’obtenir la réponse impulsionnelle (RI) du canal. Celle-ci est ensuite utilisée pour estimer la position de la source, l’angle de rasance et les pertes de transmission. Le warping conservant l’énergie, le niveau de chaque écho peut être alors estimé sur le signal déformé sans devoir revenir dans ...

Abstract-In passive acoustic monitoring, source localization using multipath propagation can be challenging whenever the source-receiver configuration leads to overlapping multipath arrival patterns. We propose a single-hydrophone model-based method with matched-impulse response (IR) processing that exploits unresolved multipath arrival patterns to locate the source in 3D, making the best use of the bathymetric information as well as the time differences of groups of arrivals. When only groups of paths are resolved, the identity of the paths constituting the groups remains unknown, preventing the use of classical localization methods. To overcome this critical limitation, we apply an approach based on the characterization of groups of paths to estimate the source location.

In passive acoustic monitoring, source localization using multipath propagation can be challenging whenever the source-receiver configuration leads to unresolved paths. Here, we propose a single-hydrophone method to estimate the range of a source based on two steps. First, we define a time deformation operator (e.g., Bonnel et al. 2014) to estimate the signal’s impulse response (IR). Each group of paths, when warped at its time of arrival, is transformed into a tone that corresponds to a peak in the IR, which can be isolated in the warped time-frequency domain, permitting a significant improvement of the time resolution of the multipath, or groups of multipath. Next, relative arrival times are used to estimate the source range and depth. When only groups of paths are resolved, the identity of the paths constituting the group remains unknown, preventing the use of classical localization methods. To overcome this critical limitation, we apply an approach based on the characterization of groups of paths to estimate source location. We evaluate the method’s performance on synthesized data, and examine its potential use on humpback whale low-frequency modulated calls using data collected off Cabo San Lucas in 2013.