Pascale Giraudet

Clusterized Mel Filter Cepstral Coefficients and Support Vector Machines for Bird Song Identification

The method we present here is a rather simple strategy for bird songs and calls classification. It builds on known and efficient technologies and ideas and must be considered as a baseline on this challenge. As we are also co-organizing this challenge, our participation aimed at defining a baseline system, with raw features, that all other participants could compare too. We did not look for optimizing each parameter of our system, and as any other participant, we conducted all the modeling and experimentation applying strictly the rules of the challenge. The method we present is dedicated to the particular setting of the challenge. It relies in particular on the fact that training signals are monolabel, i.e. only one species may be heard, while test signals are multilabeled.

Echo-Robust and Real-Time 3D Tracking of Marine Mammals Using Their Transient Calls Recorded by Hydrophones Array

Processing marine mammals (MM) signal for 3D tracking has increasing interest for the study of the behavior of endangered species. Practically, real time systems are required, and it has been observed that echoes are very common in underwater signal, and dramatically affect the localization results. Therefore, we develop a robust real-time tracking algorithm, with echo cancellation, processing wide band MM transient signal in time domain, from records of 5 omni-directional widely spaced hydrophones. The complete algorithm is tested on 25 min. of real data from NUWC & AUTEC. The attractive outlines of our method are: it runs 2 times faster than real time, requires small frequency sample data, still estimates similar trajectories than state of the art methods

Highly Defined Whale Group Tracking by Passive Acoustic Stochastic Matched Filter

In this paper, we compare two low cost time-domain tracking algorithms based on passive acoustics. The problem consists in tracking an unknown number of sperm whales (Physeter catodon). Clicks are recorded on two datasets of 20 and 25 minutes on an open-ocean widely-spaced bottom-mounted hydrophone array. The output of the method is the track(s) of the Marine Mammal(s) (MM) in 3D space and time. Firstly, we briefly review studies of the Stochastic Matched Filter (SMF) detector and its performances with a reflected click cancellation, the Teager-Kaiser-Mallat (TKM) filtering, the source separation methods and the main characteristics of MM signals. Then, we propose a real-time algorithm for MM transient call localization. We also recall the Cramer-Rao Lower Bound (CRLB) Kay (1993) and the confidence ellipses theory to predict the reachable accuracy and compare it to the tracking results. In Section 3 we show and compare results of track estimates with results from specialized teams and compare SMF versus TKM localization. Then, the system is evaluated with the confidence ellipses on the trajectories. Finally, we discuss on the possible dynamic behavior of the whale that these localizations offer, like hunting and foraging strategies. This paper deals with the 3D tracking of MM using a widely-spaced bottom-mounted hydrophone array in deep water. It focuses on sperm whale clicks. There were previous algorithms developed in the state of the art Giraudet & Glotin (2006a;b); Morrissey et al. (2006); Nosal & Frazer (2006) but none of them has satisfying results for multiple tracks and most of them are far from being real-time. Our main goal is to build a robust and real-time tracking model, despite ocean noise, multiple reflected clicks, imprecise sound speed profiles, an unknown number ofMM, and the non-linear time-frequency structure ofmostMM signals. Background ocean noise results from the addition of several noises: sea state, biological noises, ship noise and molecular turbulence. Propagation characteristics from an acoustic source to an array of hydrophones include multipath effects (and reverberations, Fig. 1), which create secondary peaks in the Cross-Correlation (CC) function that the generalized CC methods cannot eliminate. In Caudal & Glotin (2008b); Glotin et al. (2008), we gave an extension of Giraudet & Glotin (2006b) that shows multiple tracking using TKM. Here we improve this model using SMF which also allows an efficient Inter-Click-Interval and reflected click removal process. Highly Defined Whale Group Tracking by Passive Acoustic Stochastic Matched Filter 28

Approximation of linear discriminant analysis for word dependent visual features selection

To automatically determine a set of keywords that describes the content of a given image is a difficult problem, because of (i) the huge dimension number of the visual space and (ii) the unsolved object segmentation problem. Therefore, in order to solve matter (i), we present a novel method based on an Approximation of Linear Discriminant Analysis (ALDA) from the theoretical and practical point of view. Application of ALDA is more generic than usual LDA because it doesn’t require explicit class labelling of each training sample, and however allows efficient estimation of the visual features discrimination power. This is particularly interesting because of (ii) and the expensive manually object segmentation and labelling tasks on large visual database. In first step of ALDA, for each word wk, the train set is split in two, according if images are labelled or not by wk. Then, under weak assumptions, we show theoretically that Between and Within variances of these two sets are giving good estimates of the best discriminative features for wk. Experimentations are conducted on COREL database, showing an efficient word adaptive feature selection, and a great enhancement (+37%) of an image Hierarchical Ascendant Classification (HAC) for which ALDA saves also computational cost reducing by 90% the visual features space.

Strategy for a statistical analysis of odour influence on the mammalian olfactory bulb responsiveness

Abstract This paper proposes a global strategy for statistical analysis of odour influence on the responsiveness of the mammalian olfactory bulb, the first relay of the olfactory pathway. Experiments were performed on 86 mitral cells recorded in 17 anaesthetized freely breathing rats. Five pure odours and their binary mixture were used. The spontaneous activity and odour-evoked responses of the cells were characterized by their temporal distribution of activity along the respiratory cycle, i.e. by cycle-triggered histograms. Several statistical analyses were performed to describe the influence of binary odour mixtures and, especially, to detect a possible dominance of one component of the mixture.