Developing a Voice Control System for a Wheeled Robot

In recent years, domestic robots have become more functional, leading to their integration in peoples’ daily routines. However, most users are not experienced enough in human–robot interaction, necessitating simplified interfaces to bridge this gap. One approach involves using natural language as an intuitive form of communication. Insofar as using natural language doesn`t require any special skills, it makes robot control easier for non-experts. The first section of this paper includes an overview of voice-control work to-date, with references to state-of-the-art approaches. The second section proposes a hybrid architecture for a voice-based interface, combining machine learning techniques and rule-based methods. This approach reaches 95.4% and 98.8% accuracy on a small and larger model in the case of using clear speech and 88.7% and 90.3% for mumbled speech.