Auditory Object Recognition: An Attentive Neuro-Computational Approach

Опубликовано 8 августа 2016, 19:20

The talk will present work on auditory object recognition based on a neuro-computational representation model. Humans are not only adept at recognizing various auditory objects but can also to do it robustly in the presence of various distractions. Moreover, humans are able to adapt this recognition process to different tasks and demands of top-down modulatory attention. In this talk, we explore questions of this nature which are of great interest to both engineering and perceptual sciences. This work aims at 1) developing a neuro-computational framework for recognition of auditory objects based on neurophysiological processing of auditory stimuli, 2) investigating the role of goal-directed feedback mechanisms in modulating the efficacy of this recognition process. The former is useful for a wide range of tasks like musical instrument recognition, scene classification, speech recognition, etc. The later aims at giving insight into mechanisms of attention and their role in adapting sensory processing to tasks or goals and changing acoustic environments.