Learning and Intelligent Systems: Learning Binaurally-Directed Movement

9720334 Krishnaprasad本研究项目的目标是研究中枢神经系统中的时间编码，特别是谷仓猫头鹰的听觉系统，这种编码的早期发展，相关映射的学习，以及这种声音编码和映射在来源定位和声音分离中的开发。该方法包括电生理和解剖学研究，再加上神经电路的数学建模，对相关学习算法的结构和性能进行了严格的研究，并建立了一个实验机器人试验台。这个试验台是一个双耳头部，旨在能够通过纯听觉伺服，通过开发能够学习机器人听觉空间地图的控制体系结构，并利用对谷仓猫头鹰听觉系统的不断发展的理解，能够将自己定位到复杂的声学环境中的声源。这项研究的结果将为自主机器人系统中听觉感知、解释和辨别的新角色的设计提供见解。这项研究可能会导致在声音杂乱的环境中进行免提的人机通信，以及在高度自动化的制造工厂等复杂环境中进行监控。

9720334 Krishnaprasad The goal of this research project is to investigate time coding in the central nervous system, specifically the auditory system of the barn owl, the early development of such codes, the learning of associated maps, and the exploitation of such sound codes and maps in source localization and sound separation. The approach consists of electrophysiological and anatomical study, coupled with mathematical modeling of neural circuitry, the rigorous investigation of the structure and performance of relevant learning algorithms and the creation of an experimental robotic testbed. This testbed, a binaural head, is intended to be capable of orienting itself to sound sources in complex acoustic environments through pure auditory servoing, by utilizing the development of control architectures capable of learning maps of the auditory space of the robot, and drawing upon an evolving understanding of barn owl auditory system. The results of this research will provide insights into the design of novel roles for auditory sensing, interpretation and discrimination in autonomous robotic systems. This research could lead to applications in hands-free human-machine communications in acoustically cluttered environments and in monitoring complex environments such as highly automated manufacturing plants.