nIrVAna

There is mounting evidence that variability has an important role in the outcome of sensorimotor learning: in some cases, increased variability during learning is eventually associated with better performance. Regarding developmental vocal learning, an important component of human language, a basal ganglia (BG)-cortical circuit might control variability facilitating the acquisition and maintenance of vocal traits. With this project, we propose to 1) investigate the effects of variability on developmental vocal learning using birdsong as a biological model and 2) to use birdsong vocal trajectories and BG-cortical neuronal dynamics to build computational models. Taking advantage of recent methodological advances including the successful use of a robot bird as a song tutor and new chronic electrophysiological recording system developed and adapted for use in awake singing birds, we will track the song development of young zebra finches (Taeniopygia guttata) and canaries (Serinus canaria) exposed to different song models that exhibit significant variability. Then, these results will help building new developmental algorithms for neurorobotics including direct model learning and sensorimotor skill acquisition. In particular, the song learning process will be reproduced in a robot bird that will learn the correspondence between the sound emitted by a live bird and its own motor primitives This interdisciplinary project combines neuroethology, neurorobotics and signal analysis and synthesis. Ultimately, these results could help developing tools needed to implement strategies that improve the quality of learning in birds and to some extent to human infants.