A Novel Process Meta-model for Developing Automatic Speech Recognition Systems

Nowadays, Automatic Speech Recognition (ASR) figures as the main character in the Human-Computer Interaction (HCI) scenario, not only for its great usability but also for its recent advances. Several commercial applications have been launched with built-in speech interfaces. Yet, building a speech recognition system is not a trivial task. Although the speech recognition disciplines are very well documented, they occur in an unrelated way. It is difficult to understand the logic process for building such a system. We propose a development process meta-model for ASR systems in order to guide anyone to build their own ASR system. The process is validated with two distinct case studies: a context-sensitive ASR for navigating mobile robots and a semantic-driven ASR for controlling a home automation system.