Automatic speech recognition to teleoperate a robot via Web

Abstract

This article shows the way a speech recognition module has been implemented that provides the translation between voice input and the convenient text based commands to be executed on an already existing web robot. The novel contribution is the way the procedure is defined to be run over the Internet, and the interface implemented to connect any kind of external speech recognition program to the robot controller. First section gives an overview to the whole project, from a user point of view. Secondly, the software architecture is presented, giving low level details on the connectivity procedure and the interface defined between the remote controller and the speech recognition and synthesizer program. Moreover, it introduces some of the technologies (e.g. Microsoft Speech SDK) used to implement this capability. Finally, results are shown by means of both, accuracy details of the speech recognition procedure, and computing time.