Séminaire de Mathieu BERNARD de l'Institut des Systèmes Intelligents et Robotiques, Université Pierre et Marie Curie organisé par l'INRIA Montbonnot

Vendredi 25 novembre 2011 à 14h, INRIA Rhône-Alpes (room F107).

Mathieu Bernard will present a new method for self-supervised sensorimotor learning of sound source localization. Based on intensity cues computed by a bio-inspired binaural hearing system, this method allows a simulated listener to learn an auditori-motor map from the sensori-motor experience provided by an auditory evoked behavior.

The map represents the auditory space and is used to estimate the azimuthal direction of sound sources. The learning mainly consists in non-linear dimensionality reduction of sensory and motor data. Our results show that an auditori-motor map can be successfully learned, providing accurate source direction estimations. The method presented here should address promising applications in the field of active perception in autonomous robotics.
In a second time, he will present a model for supramodal audio-tactile texture discrimination. This model, proposed during the development of artificial whiskers for the robot-rat Psikharpax, is based on the observation that auditory and tactile transduction share similarities and that there exist strong interactions between auditory and tactile spectral processing within the human auditory cortex. This results clearly show the ability of the model for texture recognition in both auditory and tactile tuning