Novice User Experiences with a Voice-Enabled Human-Robot Interaction Tool

摘要

Voice recognition software is a widely adopted tool in a variety of task domains. However, several mission critical systems, which have high security demands cannot allow outside connections to the remote systems that provide voice recognition capabilities. This presents a problem for modern day voice recognition, which is largely cloud based. To address this issue, we leveraged Julius as an offline phoneme-based voice recognizer in order to incorporate voice recognition software into robotic systems for law enforcement officers. In order to address the difficulties that officers with a variety of dialects have when interacting with a phoneme-based voice recognizer, a training tool was developed. This paper examines the lessons learned from the latest implementation of the training tool over the course of several voice-enabled Human-Robot Interaction (HRI) experiments. The majority of these users were novices who had little to no experience with voice recognition software. Interactions were completed at three events in Ko?ice, Slovakia: (1) Museum Night 2018, (2) a private company demonstration, and (3) Technical University of Ko?ice's Summer Kids University (TUKE for kids) demonstration. The results of the user interaction evaluations highlighted that, through training, novice users could learn to interact with an offline voice recognition system after a short period of time by operating a simulated robotic system.