Towards an Ambient Intelligent Environment for Multimodal Human Computer Interactions

摘要

Recent years have seen an explosion in the academic and commercial applications of digital assistants. These technologies have become increasingly prolific, and their use has resulted in fairly rigid and standardized techniques for achieving a desired result. Whether this includes physical actions, or direct voice queries to the system via predefined wake words and queries, there is a stark boundary between the human and the system. We aim to explore a shift in the paradigm of these current implementations, to that of an Ambient Intelligent (AmI) environment in which users can interface with the system in a more natural, seamless, and multi-modal manner. Applications of this type of technology range from assisted living, to smart conference rooms and meeting spaces. In this paper we introduce an architectural framework for building an ambient intelligent platform using a combination of video and audio sensors to capture and process the data in a given area of interest.