Robot Operation via Natural Voice Control

In this paper we present an introductory implementation of a natural language movement input system for the BWI robots. The objective is to implement simple and intuitive voice commands as a basic interface for human-robot interaction. The need for this interface arises from the desire to eventually create autonomous, intelligent, robots that can perform a multitude of tasks. These robots have to be capable of carrying out orders from humans through some form of input. Besides the most common input source, vision, audio presents a broad interface that can be used to gather information from the environment. We will present a simple proof-of-concept of such an interface. This interface is built through the use of pocketsphinx, an open-source library created by Carnegie Mellon University that works as a speech recognizer. Phrases are parsed into basic commands (a command language that we have defined for the sole purpose of this endeavor) that the robot can then execute.